Pseudo Color Image Processing on X-ray images, NV images
Content
PSEUDO COLOR IMAGE PROCESSING
A Minor Project Report Submitted in Partial Fulfillment of the Requirements for the Degree of
Bachelor OF TECHNOLOGY IN ELECTRONICS & COMMUNICATION ENGINEERING
By Jaydip R. Fadadu 08BEC024 Kuldip R. Gor 08BEC030 Under the Guidance of Prof. Tanish H. Zaveri
Department of Electrical Engineering Electronics & Communication Engineering Program Institute of Technology, Nirma University Ahmedabad-382481 November 2011
CERTIFICATE
This is to certify that the Minor Project Report entitled “ Pseudo Color Image Processing “ submitted by Jaydip Fadadu (Roll No. 08BEC024) and Kuldip Gor ( Roll No. 08BEC030) as the partial fulfillment of the requirements for the award of the degree of Bachelor of Technology in Electronics & Communication Engineering, Institute of Technology, Nirma University is the record of work carried out by his/her under my supervision and guidance. The work submitted in our opinion has reached a level required for being accepted for the examination.
Date:
Prof. Tanish H. Zaveri Project Guide
Prof. A. S. Ranade HOD (Electrical Engineering) Nirma University, Ahmedabad
ACKNOWLEDGEMENT
It gives us great pleasure in expressing thanks and profound gratitude to we like to give our special thanks to our Faculty Guide, Prof Tanish Zaveri, Professor, Department of Electronics & Communication Engineering, Institute of Technology, Nirma University, Ahmedabad for his valuable guidance and continual encouragement throughout the Minor Project. We are heartily thankful to him for his regular suggestions and the clarity of the concepts of the topic that helped us a lot during this project. We are also thankful to Prof A. S. Ranade, HOD, Department of Electronics & Communication Engineering, Institute of Technology, Nirma University, Ahmedabad for his continual kind words of encouragement and motivation throughout the Minor Project. We are also thankful to Dr K Kotecha, Director, Institute of Technology for his kind support in all respect during our work. We are thankful to all faculty members of Department of Electronics & Communication Engineering, Institute of Technology, Nirma University, Ahmedabad for their special attention and suggestions towards the project work. The friends, who always bear and motivate us throughout this course, we are thankful to them.
Mr. Jaydip Fadadu 08BEC024
Mr. Kuldip Gor 08BEC030
ABSTRACT
Human eye can distinguish only a limited number of gray scale value but can distinguish between thousands of color. So it is clear that human can extract more amount of information from the colored image than that of the gray image. So pseudo coloring is very useful in improving the visibility of an image. Certain application like buggage scanning at airport or in medical field or in night vision camera works in the band of X ray and infrared so the images produced by this techniques always gives gray image. At the same time the proper inspection of this all images is very crucial and critical. So coloring of those images is very important. Moreover color should be applied in such a way that it improves the visibility of image in optimal way for that particular application. The project discusses various techniques of pseudo coloring for images of above mentioned applications.
Index
Chapter No. Title Acknowledgement Abstract Index List of Figures List of Tables Nomenclature 1 Introduction 1.1 Introduction 1.2 1.3 1.4 2 3 Necessity of project Objective of project Contents of the report Page No. i Ii iii iv v vi 1 1 1 2 3 4 7 8 8 9 9
Classification of Pseudo Coloring Technique Pseudo Coloring of Night Vision Images 3.1 False Color Fusion and Color Enhancement 3.2 3.3 3.4 Color Based Clustering Using Hill Climbing Cluster Recognization Color Transferring
4
Pseudo Coloring of x-ray Images of Medical & 11 Luggage scanning @ Air-port 4.1 Simple Process of Pseudo Coloring 11 4.2 4.3 Various Coloring Techniques 12 Proposed Method based on Look Up Table 18 designed from Warm Color Scale 23 Evaluation Parameters Conclusions 24
5
References
25
LIST OF FIGURES
Fig. No. Title Page No. 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Block Diagram Of Color Transfer Method Results of Night Vision Color Image Simple Block Diagram of Pseudo coloring The HOT color Scale R,G and B v/s I for HOT coloring The JET color Scale R,G and B v/s I for JET coloring R,G and B v/s I for HSI coloring The HSI color Scale based on Concave Part(solid part of Fig. 8) The HSI color Scale based on Convex Part(solid part of Fig. 8) The RAINBOW color Scale R,G and B v/s I for RAINBOW coloring Simulation Results of various Coloring Methods
7 10 11 12 13 13 13 14 14 14 15 15 16
Detailed Block Diagram of Proposed Method based on Look Up 18 Table designed from Warm Color Scale
15 16 17
The WARM color scale R,G and B v/s I for WARM coloring
20 20
Simulation Results of Proposed method based on Look up table 21 created as per warm color scale
LIST OF TABLES
1
Entropy and Colorfullness Mertic Comparision
23
NOMENCLATURE
Vis IR Visible Image Infrared Image Mean kth cluster in source images Standard deviation of kth cluster in source images Mean of nth target image Standard deviation of nth target image Q Similarity matrix
RGB Red-Green-Blue Color space HSI Hue-Saturation-Intsnsity Color space
Chapter 1 Introduction
1.1 Introduction
Pseudo-color image processing assigns color to grayscale images. This is useful because the human eye can distinguish between millions of colors but relatively few shades of gray. Pseudo-coloring has many applications on images from devices capturing light outside the visible spectrum, for example, infrared and X-rays. A pseudo-color image is derived from a grey scale image by mapping each pixel value to a color according to a table or function. The table or the function is decided in such a way that it enhances the visibility optimally for a particular application. To achieve that objective various color scales are defined in the literature. The color scales are defined in different color spaces like RGB or HSI etc. before performing the actual color assigning the gray scale image is enhanced by performing certain operations on it. The project discusses the various techniques of enhancing the image and performing pseudo coloring on it to enhance the visibility.
1.2 Necessity Of the Project
Achieving higher threat detection rates during inspection of X-ray in luggage scans and medical field is a pressing and sought after goal for airport and airplane security personnel as well as doctors. Because of the complexities in knowing the content of each individual bag and the increasingly sophisticated methods adopted by terrorists in concealing threat objects, X-ray luggage scans obtained directly from major luggage inspection machines still do not reveal 100% of objects of interest, especially low density threats. So in both the cases if the gray density value of the threat is very low then it may easily go unchecked. Various types of grayscale enhancements are done to improve the detection but It was shown, through screener evaluation studies that an improvement of 57% in detection rates accrued after enhancement as compared to the inspection results from the raw data. Furthermore, since it is known that human beings can only discern a few dozen gray level values while they can distinguish thousands of colors, the use of color for human interpretation could only improve the number of objects that can be distinguished. In addition, color adds vivacity to images, which in turn decreases boredom and improves the attention span of screeners. Moreover Modern night vision camera systems are used for military and law enforcement applications to design
intelligent surveillance systems for security. These systems are designed to provide enhanced image with better perceptual quality in adverse environmental conditions. The most common night-time imagery systems capture images in two spectral bands, near infrared (NIR) and visual, thus providing complementary information of the observed scene which enables the observer to perceive more complete picture of the scene with a larger degree of situational awareness. A fused image combines all the salient information from the source images which is more suitable for human/ machine perception. Image fusion is an effective way of reducing the volume of information while at the same time it preserves all the useful information from the source images. The rapid development in the technology of night vision (NV) systems has led to a growing interest in the natural color display of night vision imagery. As human visual system is more sensitive to color information, efficient color transfer methods are required to enhance color image which has several benefits over gray image. The color transfer methods improve feature contrast, allowing better scene recognition and object detection which is useful in surveillance, reconnaissance, and security applications.
1.3 Objective Of the Project
Given the fact that the human eye is more sensitive to some parts of the visible spectrum of light than to others and that the brain may interpret color patterns differently, the interpretation of results produced by visualization techniques depends crucially on the color mapping applied. In an effort to address the relatively new problem of visualizing lowdensity threat items in x-ray images, while incorporating considerations of the perceptual and cognitive characteristics of the human operator, a set of RGB- and HSI- based color transforms were designed and applied to single energy X-ray luggage scans. The analysis showed that color coding resulted in a large improvement in the detection rates of threat objects in luggage and in an increased screener’s visual and mental alertness and attention retention. In night vision color images it is very important that coloring should be natural. It mean colr should be applied in such a way that after coloring the scene seems natural at the same time it should be efficient in identifying the human presence. So the same method cannot be applied to both night vision infrared image and X-ray image. So two very different methods are proposed in the project for pseudo coloring of both different types of images.
1.4 Content Of The Report
The report starts with the basic introduction of the pseudo coloring. Chapter 1 discusses the basic need of applying pseudo color to gray scale image and its application in night vision camera and X-ray images. Chapter 2 discusses the various types of coloring techniques that are used in different fields. Then proceeding chapter discusses the pseudo coloring of night vision images using image fusion and how to apply the natural color to NIR images. The next chapter discusses the pseudo coloring of X ray images.
Chapter 2
Classification Of Pseudo Coloring Technique
Based on available literature, pseudo coloring techniques can generally be grouped into five main categories:
1) Spectrum-based maps, where color scales are designed by having the hue sequence
range from violet, via indigo, blue, green, yellow, and orange, to red, following the color order of the visible spectrum. Since the human visual system has different sensitivities to different wavelengths, researchers such as Clarke and Leonard indicated that spectrum-based color scales were not perceived to possess a color order that corresponds to the natural order of grayscale in the image.
2) Naturally ordered maps, where a number of researchers attempted to define a
particular path traversing the RGB color space under certain predefined constraints. The heated-object scale is achieved by bringing the RGB intensities up in the order of red, green, and blue, and limiting the path to 60◦ clockwise from the red axis. This selection is based on the claim that natural color scales seem to be produced when the intensities of the three primary colors rise monotonically with the same order of magnitude throughout the entire scale. To construct a natural color scale, Lehmann et al defined a spiral-like scale in the RGB model, keeping the original brightness progression of grayscale images. Specifically speaking, their color scale follows a spiral-like path along the diagonal of the RGB model. The authors formularized such a scale to allow the determination of the resulting number of colors, and demonstrated their scale’s effectiveness by applying it to medical X-ray images. They also pointed out that better results could be obtained if other color models were used. The use of HSI was suggested as future work based on the fact that the lightness component is directly represented by one of the axes in HSI.
3) Uniformly varying maps, where several studies focused on constructing a uniform
color scale where adjacent colors are equally spaced in terms of just noticeable differences (JNDs) and maintain a natural order along this color scale. Levkowitz and Herman’s research provided a scale with the maximum uniform resolution. The authors were hoping that their optimal color scale outperforms the grayscale, but evaluation results did not confirm that, at least for their particular application. They presented several reasons that might have caused
the unexpected results. One particular reason was that they used the CIELUV model to adjust the final colors which might not have been appropriate to model the perceived uniformity. Another reason was that the perceived change in color due to its surrounding was not taken into consideration. Shi et al designed a uniform color scale by visually setting up its color series to run from inherently dark colors to inherently light colors; i.e., from black through blue, magenta, red, yellow to white, then further adjusting the colors to make them equally spaced. The color scales were evaluated by comparing them to the grayscale. The authors indicated that the contrast sensitivity has been improved after applying their uniform scale, but they failed to demonstrate any significant outcome.
4) Function-based maps, where researchers utilized common mathematical functions
such as the sine function to construct desired mappings or color scales. Gonzalez and Woods described an approach where three independent mathematical transforms were performed on the gray level data, and the three output images fed into the R, G, and B channels to produce a specific color mapping.
5) Refernce Natural Images Database Technique
There are various approaches available in literature, which are broadly divided into three different categories: statistical approach, region-based approach, pattern recognition based approach. However from the recent literature, we found that combinations of these approaches are used to improve the quality of resultant image. Yang et al. in have proposed region-based approach for color transfer in night vision image sequences. The method uses support vector machines for region recognition among a set of natural color database. In case of specific application of color transfer in night vision FLIR (Forward Looking Infrared) images based on texture pattern recognition for color transfer images, Sun and Zhao proposed a method for automatic natural color mapping for FLIR. A local-coloring method utilizing image analysis and fusion was introduced by Zheng and Essock in, which render the NV image segment-by-segment by taking advantage of image segmentation, pattern recognition, histogram matching and image fusion. Recently, A. Toet proposed a fast color mapping method in which the mapping optimizes the match between the multi-band image and the reference natural image, and yields a night vision image with a natural daytime color appearance. This lookup-table based mapping procedure is simple and provides object color constancy. However the scene matching between source and target is performed manually
which may yield less naturalistic results for images containing regions that differ significantly in their content. Gang and Huang presented a multi-scale color image fusion using contourlet transform and expectation maximization (EM) where the color transfer is implemented in YUV color space. A new linear color space ICbCr was proposed by Xu and Li in especially for multiband night vision imagery to transfer the color distribution of the target image to the source NV images.
Chapter 3
Pseudo Coloring Of Night Vision Images
In the color transfer method color based clustering is applied on color map in LAB color space and cluster recognition is performed based on color similarity metric. The block diagram of the proposed color transfer method is shown in Figure 1. As shown in Figure 1, false colored nightvision image is obtained by assigning linear combinations of visible and IR source images to the RGB channels. The false colored NV image is enhanced by decorrelation stretch and contrast stretch. A colormap of the enhanced false colored
Fig. 1 Block Diagram Of Color Transfer Method
NV image is obtained. Color based clustering is performed on the colormap in the LAB color space using hill climbing algorithm. The association between each cluster and a natural color image in the target color look-up table is carried out automatically utilizing a nearest neighbour criteria based on a color similarity metric. The color components in each index within a cluster of the colormap are modified by statistics matching with the corresponding natural color image. The natural colored nightvision image is produced from the new colormap. Finally, the natural colored NV image is transformed to HSV color space and the enhanced gray image is substituted in the “value (V)” component to generate the final output of the proposed method. The subsequent subsections of this section describe the detailed explanation of the major steps of method [3].
3.1 False Color Fusion and Color Enhancement
The false color fused RGB image can be represented by the following equations: R(m; n) = (Vis(m; n) + IR(m; n)) G(m; n) = IR(m; n) B(m; n) = Vis(m; n) - IR(m; n) The false color fused image so formed has intensity variations similar to visible and IR source images. In order to achieve better seperation in color based clustering we perform decorrelation stretch for color enhancement and linear contrast stretch for intensity enhancement. Decorrelation stretch as described in increases color seperation across highly correlated channels while keeping the band variance same. Decorrelation stretch makes many features easier to observe which were not clearly visible in the original image.
3.2 Color based Clustering using Hill Climbing
Color based clustering is performed on the LAB colormap using the hill climbing algorithm. A color based image segmentation method using hill climbing algorithm proposed by Ohashi et al. is utilized for colormap clustering in the proposed method. The number of clusters required for proper classification of colormap are automatically determined by the hill climbing algorithm. A three-dimensional histogram is computed in the LAB color space. The hill climbing algorithm can be seen as a search window being run across the space of the 3D LAB histogram to find the largest bin within that window. Each bin in the color histogram has 3d - 1 = 26 neighbours where d = 3 is the number of dimensions of the feature space. The number of peaks found indicate the value of K and the value of those bins form the initial
seeds for the K-means segmentation. Thus, local maxima are found for clusters in the 3D color histogram of the colormap in the LAB color space. The entries of the colormap are then associated with the detected local maxima to generate several coherent clusters in the LAB colormap.
3.3 Cluster Recognition
The target color look-up table is created as follows. Each image from the natural color target image database is firsth smoothed by separately applying low pass filter in each RGB channels. This operation reduces the number of colors and enables the extraction of dominant colors from the natural image. The smoothed image is then transformed into LAB color space and first order statistics, mean µ and standard deviation , are computed for each band. A nearest neighbour criteria is used for automatic association of a cluster of colormap with a unique natural color image in the target color look-up table. The similarity metric
Where
Is used for the measure of the distance between Kth cluster and nth image.
3.4 Color Transfer
Color transfer is performed cluster-by-cluster by the standard statistics matching method proposed by Toet. Each index in LAB colormap is first checked as to which cluster does it belong and then the statistics of natural color target image associated with that cluster is transferred to the LAB values of the index in colormap. Thus a new LAB colormap is obtained. The equations for color transfer on each index in the colormap are defined as follows:
Kth cluster is associated with nth natural image. deviation of kth cluster in source images. deviation of nth target image.
and and
indicates the mean and standard indicates the mean and standard
Fig. 2 Results of Night Vision Color Image
Chapter 4
Pseudo coloring on X-ray images of Medical and Luggage scanning @ Airport
In this project we mainly focus on two major areas: 1. Medical Pseudo coloring is done to enhance the visibility of the fracture or any disease which is not clearly visible by naked eyes in x-ray image. 2. Luggage Scanning @ Air-port Pseudo coloring is done to increase the detect ability of the low density weapons which are not clearly detectable in simple x-ray scanning.
4.1 Simple Process of Pseudo Coloring
To develop enhanced color image from x-ray image, processing is done on x-ray image. A simple block diagram of that process is shown in figure 3. Few steps are followed before coloring to increase the visibility.
X-Ray Image
Enhanced Color Image
Contrast Stretch
Color Conversion
Salt & Pepper Noise Removal
Fig. 3 Simple Block Diagram of Pseudo coloring
A. X-ray Image: It is the simple black and white image taken by x-ray camera. Each pixel of the image has information i.e. intensity of which value varies from 0 to 255. Processing is done on this image. B. Contrast stretch This process increases the difference between maximum and minimum intensity. In most of the methods simple linear stretch is applied. C. Salt & Pepper Noise Removal After contrast stretching, Salt and pepper noise is removed. It is done by applying various filters. In most of the cases median filter is applied. At the end of this block we get enhanced gray image with more information regarding weapons in case of luggage scanning or fractures in case of medical images. D. Color Conversion It is the process of converting enhanced gray scale image into color image. Various methods for color conversion are available [1][2][5]. We have studied 4 basic methods: i. ii. iii. iv. Hot Jet HSI Rainbow
All the methods are further explained in detail.
4.2 Various Coloring Techniques
1. HOT It is RGB based linear color map. “Hot” changes smoothly from black, through shades of red, orange, and yellow to white as shown in fig. 4.
Fig.4 The HOT color Scale In fig. 3 RGB values versus gray value is shown for HOT coloring. That plot shows the variation of R,G and B with respect to I.
Fig. 5 R,G and B v/s I for HOT coloring
2. JET It is RGB based linear color map. It ranges from Blue to Red, passing through Cyan, Yellow and Orange as shown in fig.4. In fig. 5 RGB values versus gray value is shown for JET coloring. This color map can be obtained by converting this plot into simple mathematical formulas.
Fig. 6 The JET color scale
Fig. 7 R,G and B v/s I for JET coloring
3. HSI It is Histogram based non-linear color mapping. The colors of the various components in the scene are assigned based on the values of the raw image. Pixel ranges are selected from the data’s histogram and automatically given certain colors. For example, four gray-level regions were created, the chances of the low density threat being present would be greatest in the first two regions. Blue will be used as background and other easy-to-remember basic colors like red and green are applied to the other pixels in each bin. The output image would have four hues, which vary as a function of the gray intensity value of each pixel.
Fig. 8 H,S and I v/s gray for HSI based coloring
Fig. 9 HSI color scale based on concave part(solid part of fig. 8)
Fig. 10 HSI color scale based on convex part(solid part of fig. 8)
4. Rainbow It can be considered as a special case of Sine/Cosine transform. 3 color transfer functions are used for rainbow map. All the functions are periodic in the sense that they get peak in particular color interval. Rainbow based color map and RGB’s relation with I are shown in following figures.
Fig. 11 The Rainbow color scale
Fig. 12 R,G and B v/s I for Rainbow color map
Fig. 13 Simulation Results of various Coloring Methods
4.3 Proposed Method based on Look Up Table designed from Warm Color Scale
This method is specially designed for pseudo coloring in x-ray images for weapon detection and medical. In fig. 14 a detailed block diagram is shown.
X-Ray Image Warm Color Map Adaptive Histogram Equalization
Look Up Table Contrast Stretch Using Intensity Adjust
Noise Removal Using 2D-Median Filter
Enhanced Gray Scale Image
Color Conversion Using Look Up Table
Enhanced Color Image
Fig. 14 Detailed Block Diagram of Proposed Method based on Look Up Table designed from Warm Color Scale
Initially Look Up Table is created based on WARM color scale. Warm color scale is explained in [1][2][5]. A simple x-ray image is passed through various blocks to enhance the information regarding fractures/cracks or weapons in case of medical or weapon detection @ airport respectively. Pseudo coloring on this enhanced gray image is done based on the Look Up Table prepared earlier as per warm color scale.
Explanation of the above block diagram: [1]. X-Ray Image It is the simple black and white image taken by x-ray camera. It may be the image of any body part taken for medical use or may be image of weapon detection @ air-port. Each pixel of the image has information i.e. intensity of which value varies from 0 to 255. Processing is done on this image. [2]. Adaptive Histogram Equalization It is the process of enhancing the contrast of images by transforming the values in the intensity image I. Unlike simple histogram equalization it operates on small data regions (tiles), rather than the entire image. Each tile's contrast is enhanced, so that the histogram of the output region approximately matches the specified histogram. The neighboring tiles are then combined using bilinear interpolation in order to eliminate artificially induced boundaries. The contrast, especially in homogeneous areas, can be limited in order to avoid amplifying the noise which might be present in the image. In this proposed method tile size is taken as 16 x 16. [3]. Contrast Stretch using Intensity Adjust By this block linear intensity contrast stretch is applied. It increases the difference between maximum and minimum intensity. MATLAB inbuilt function imadjust is used for this function. [4]. Noise Removal Using 2D-Median Filter. At the end of the contrast stretching we get the image which may have few unwanted noise dots. It should be removed for false detection. This can be removed by filter. In this proposed method 2 dimensional median filter is used. It is the filter which takes the median of given square block. Here we have taken 3 x 3 block.
[5]. Enhanced Gray Scale Image This is the final gray scale image i.e. enhanced image having more information than original one. Now the coloring is done on this image. [6]. Warm color scale This is non-liner color scale. It varies from Dark Blue to Light Yellow through Magenta and Orange as shown in fig. 15. The distances of adjacent colors on this scale are perceivably equal. A 256-step scale as seen in Fig. 11 was developed. For any intensity I(i) and I(i+1) the correspondence (R, G, B)(i) < (R, G, B)(i+1). This law is followed throughout the color scale. Respective relation of R,G and B with I is shown in fig. 16.
Fig. 15 The WARM Color Scale
Fig.16 R,G and B v/s I for WARM coloring
[7]. Look Up Table It is a simple table which gives the corresponding R, G and B values for given intensity I. It is designed based on the Warm Color Scale which is already explained earlier.
[8]. Color Conversion using Look Up Table Color Conversion i.e. I to (R,G,B) is done based on the look up table created earlier. The coloring is done on enhanced gray scale image. It is just a simple assignment of (R,G,B) as per the value of intensity of that pixel.
[9]. Enhanced Color Image This is the final color image having more, clear information regarding fracture or cracks in medical images or weapons in luggage scanning @ airport.
Fig. 17 Simulation Results of Proposed method based on Look up table created as per warm color scale
Chapter 5
Evaluation Parameters for Coloring
The quality assessment of different image fusion schemes for X-ray images is traditionally carried out by subjective evaluations. The subjective evaluation is influenced by individual human perception. In recent literature [4] objective evaluation parameters are proposed. Here we have considered non reference based evaluation parameters; entropy and colorfulness metric. Entropy is used to measure the information content of an image. The entropy of a grayscale image is:
where G is the number of gray levels in the image’s histogram (which can be 255 for a typical 8-bit image) and p(i) is the normalized frequency of occurrence for each gray level, i.e., the histogram of the image. The entropy of colored image is computed for each band in RGB color space and average of the entropy of the three bands is considered for evaluation. Colorfulness metric is an efficient metric for calculating colorfulness of images and it is described in [4]. Larger the color variations in the image, higher is the colorfulness metric. The proposed algorithm is compared with the standard statistics matching method proposed by Toet. The simulation results of proposed algorithm are shown in Figure 17 and the simulation results of other coloring techniques are shown in figure 13. It is observed that proposed method provides more informative appearance compared to other method [1] and [5] as shown in Figure 13 and Figure 17. Hence the cracks in medical image and weapons in luggage scanning the scene can be easily distinguished. Their comparison with other techniques using parameters like entropy and colorfullness metric are shown in table 1.
ENTROPY
NO
1 2 3 4 5 6
HOT
0.9734 0.6764 0.7389 0.7643 0.6358 1.5171
JET
3.607 3.5149 3.6231 3.8605 3.5244 3.6661
HIS
5.4859 5.5525 5.5841 5.5473 5.7071 5.4594
RAINBOW PROPOSED
6.4439 6.4578 6.5618 6.443 6.9876 5.9877 5.772 5.5401 5.5839 5.5479 5.6924 5.7758
COLORFULLNESS METRIC
HOT
1 2 3 4 5 6 165.82 159.18 171.64 185.31 156.77 210.2
JET
0.8937 0.8947 0.9241 0.9396 0.811 1.073
HIS
0.989 0.9973 0.8892 0.9736 0.7456 0.8602
RAINBOW PROPOSED
0.3512 0.3512 0.3512 0.3512 0.3512 0.3512 76.3414 73.7114 73.7709 81.549 78.3135 87.2228
Table 1. Entropy and Colorfullness Mertic Comparision
CONCLUSION
Thus in this report we presented efficient method for pseudo coloring of grayscale images. Here we discussed the pseudo coloring of two different types of images. The first is Night vision infrared image and second is X ray images. Night vision images are used in army and nevy applications while the X ray images are used at air ports as well as in medical. Thus we can see that it is very crucial that the image should be visibly best as possible. So that personnel can extract information as much as possible. So we apply color to those images in this project. We can see that the colored images are far better to analyze than the gray images. Moreover same method cannot be applied to two different type of the field. So we used two very different methods to for night vision images and X ray images. For night vision images we try to apply natural color to images so that soldiers can relate the image to the surround environment and looks familiar. So we prepared the database using natural reference image and applied color transfer using that. In x ray we used the applied the standard color scales defined in the literature and applied. We also developed a color look up table for coloring and we observe that in this method we obtained the best result. In future we still find the another method for colorization. The pre-processing on the grayscale images is very crucial and more enhancing algorithms can be applied to obtain still better results.
References
[1]. Andreas Koschan and Mongi Abidi, "Digital Color Image Processing," A John Willy & Sons, INC., Publication, Hoboken, New Jersey. [2]. Rafael C. Gonzalez and Rechard E. Woods, “Digital Image Processing” Prentice Hall, New Jersey. [3]. Tanish Zaveri, Mukesh Zaveri, Ishit Makwana and Harshit Mehta,“ An Optimized Region-based Color Transfer Method for Night Vision Application”. [4]. [5]. Toet. Natural color mapping for multiband nightvision imagenary. Information Fusion, vol. 4(3), pp. 155-166, 2003. Besma R. Abidi, Senior Member, IEEE, Yue Zheng, Andrei V. Gribok, and Mo ngi A. Abidi, Member, IEEE. “Improving Weapon Detection in Single Energy X-Ray Images Through Pseudocoloring” Ieee transactions on systems, man, and cybernetics—part c: applications and Reviews, vol. 36, no. 6, pp. 784-796, November 2006.
A Minor Project Report Submitted in Partial Fulfillment of the Requirements for the Degree of
Bachelor OF TECHNOLOGY IN ELECTRONICS & COMMUNICATION ENGINEERING
By Jaydip R. Fadadu 08BEC024 Kuldip R. Gor 08BEC030 Under the Guidance of Prof. Tanish H. Zaveri
Department of Electrical Engineering Electronics & Communication Engineering Program Institute of Technology, Nirma University Ahmedabad-382481 November 2011
CERTIFICATE
This is to certify that the Minor Project Report entitled “ Pseudo Color Image Processing “ submitted by Jaydip Fadadu (Roll No. 08BEC024) and Kuldip Gor ( Roll No. 08BEC030) as the partial fulfillment of the requirements for the award of the degree of Bachelor of Technology in Electronics & Communication Engineering, Institute of Technology, Nirma University is the record of work carried out by his/her under my supervision and guidance. The work submitted in our opinion has reached a level required for being accepted for the examination.
Date:
Prof. Tanish H. Zaveri Project Guide
Prof. A. S. Ranade HOD (Electrical Engineering) Nirma University, Ahmedabad
ACKNOWLEDGEMENT
It gives us great pleasure in expressing thanks and profound gratitude to we like to give our special thanks to our Faculty Guide, Prof Tanish Zaveri, Professor, Department of Electronics & Communication Engineering, Institute of Technology, Nirma University, Ahmedabad for his valuable guidance and continual encouragement throughout the Minor Project. We are heartily thankful to him for his regular suggestions and the clarity of the concepts of the topic that helped us a lot during this project. We are also thankful to Prof A. S. Ranade, HOD, Department of Electronics & Communication Engineering, Institute of Technology, Nirma University, Ahmedabad for his continual kind words of encouragement and motivation throughout the Minor Project. We are also thankful to Dr K Kotecha, Director, Institute of Technology for his kind support in all respect during our work. We are thankful to all faculty members of Department of Electronics & Communication Engineering, Institute of Technology, Nirma University, Ahmedabad for their special attention and suggestions towards the project work. The friends, who always bear and motivate us throughout this course, we are thankful to them.
Mr. Jaydip Fadadu 08BEC024
Mr. Kuldip Gor 08BEC030
ABSTRACT
Human eye can distinguish only a limited number of gray scale value but can distinguish between thousands of color. So it is clear that human can extract more amount of information from the colored image than that of the gray image. So pseudo coloring is very useful in improving the visibility of an image. Certain application like buggage scanning at airport or in medical field or in night vision camera works in the band of X ray and infrared so the images produced by this techniques always gives gray image. At the same time the proper inspection of this all images is very crucial and critical. So coloring of those images is very important. Moreover color should be applied in such a way that it improves the visibility of image in optimal way for that particular application. The project discusses various techniques of pseudo coloring for images of above mentioned applications.
Index
Chapter No. Title Acknowledgement Abstract Index List of Figures List of Tables Nomenclature 1 Introduction 1.1 Introduction 1.2 1.3 1.4 2 3 Necessity of project Objective of project Contents of the report Page No. i Ii iii iv v vi 1 1 1 2 3 4 7 8 8 9 9
Classification of Pseudo Coloring Technique Pseudo Coloring of Night Vision Images 3.1 False Color Fusion and Color Enhancement 3.2 3.3 3.4 Color Based Clustering Using Hill Climbing Cluster Recognization Color Transferring
4
Pseudo Coloring of x-ray Images of Medical & 11 Luggage scanning @ Air-port 4.1 Simple Process of Pseudo Coloring 11 4.2 4.3 Various Coloring Techniques 12 Proposed Method based on Look Up Table 18 designed from Warm Color Scale 23 Evaluation Parameters Conclusions 24
5
References
25
LIST OF FIGURES
Fig. No. Title Page No. 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Block Diagram Of Color Transfer Method Results of Night Vision Color Image Simple Block Diagram of Pseudo coloring The HOT color Scale R,G and B v/s I for HOT coloring The JET color Scale R,G and B v/s I for JET coloring R,G and B v/s I for HSI coloring The HSI color Scale based on Concave Part(solid part of Fig. 8) The HSI color Scale based on Convex Part(solid part of Fig. 8) The RAINBOW color Scale R,G and B v/s I for RAINBOW coloring Simulation Results of various Coloring Methods
7 10 11 12 13 13 13 14 14 14 15 15 16
Detailed Block Diagram of Proposed Method based on Look Up 18 Table designed from Warm Color Scale
15 16 17
The WARM color scale R,G and B v/s I for WARM coloring
20 20
Simulation Results of Proposed method based on Look up table 21 created as per warm color scale
LIST OF TABLES
1
Entropy and Colorfullness Mertic Comparision
23
NOMENCLATURE
Vis IR Visible Image Infrared Image Mean kth cluster in source images Standard deviation of kth cluster in source images Mean of nth target image Standard deviation of nth target image Q Similarity matrix
RGB Red-Green-Blue Color space HSI Hue-Saturation-Intsnsity Color space
Chapter 1 Introduction
1.1 Introduction
Pseudo-color image processing assigns color to grayscale images. This is useful because the human eye can distinguish between millions of colors but relatively few shades of gray. Pseudo-coloring has many applications on images from devices capturing light outside the visible spectrum, for example, infrared and X-rays. A pseudo-color image is derived from a grey scale image by mapping each pixel value to a color according to a table or function. The table or the function is decided in such a way that it enhances the visibility optimally for a particular application. To achieve that objective various color scales are defined in the literature. The color scales are defined in different color spaces like RGB or HSI etc. before performing the actual color assigning the gray scale image is enhanced by performing certain operations on it. The project discusses the various techniques of enhancing the image and performing pseudo coloring on it to enhance the visibility.
1.2 Necessity Of the Project
Achieving higher threat detection rates during inspection of X-ray in luggage scans and medical field is a pressing and sought after goal for airport and airplane security personnel as well as doctors. Because of the complexities in knowing the content of each individual bag and the increasingly sophisticated methods adopted by terrorists in concealing threat objects, X-ray luggage scans obtained directly from major luggage inspection machines still do not reveal 100% of objects of interest, especially low density threats. So in both the cases if the gray density value of the threat is very low then it may easily go unchecked. Various types of grayscale enhancements are done to improve the detection but It was shown, through screener evaluation studies that an improvement of 57% in detection rates accrued after enhancement as compared to the inspection results from the raw data. Furthermore, since it is known that human beings can only discern a few dozen gray level values while they can distinguish thousands of colors, the use of color for human interpretation could only improve the number of objects that can be distinguished. In addition, color adds vivacity to images, which in turn decreases boredom and improves the attention span of screeners. Moreover Modern night vision camera systems are used for military and law enforcement applications to design
intelligent surveillance systems for security. These systems are designed to provide enhanced image with better perceptual quality in adverse environmental conditions. The most common night-time imagery systems capture images in two spectral bands, near infrared (NIR) and visual, thus providing complementary information of the observed scene which enables the observer to perceive more complete picture of the scene with a larger degree of situational awareness. A fused image combines all the salient information from the source images which is more suitable for human/ machine perception. Image fusion is an effective way of reducing the volume of information while at the same time it preserves all the useful information from the source images. The rapid development in the technology of night vision (NV) systems has led to a growing interest in the natural color display of night vision imagery. As human visual system is more sensitive to color information, efficient color transfer methods are required to enhance color image which has several benefits over gray image. The color transfer methods improve feature contrast, allowing better scene recognition and object detection which is useful in surveillance, reconnaissance, and security applications.
1.3 Objective Of the Project
Given the fact that the human eye is more sensitive to some parts of the visible spectrum of light than to others and that the brain may interpret color patterns differently, the interpretation of results produced by visualization techniques depends crucially on the color mapping applied. In an effort to address the relatively new problem of visualizing lowdensity threat items in x-ray images, while incorporating considerations of the perceptual and cognitive characteristics of the human operator, a set of RGB- and HSI- based color transforms were designed and applied to single energy X-ray luggage scans. The analysis showed that color coding resulted in a large improvement in the detection rates of threat objects in luggage and in an increased screener’s visual and mental alertness and attention retention. In night vision color images it is very important that coloring should be natural. It mean colr should be applied in such a way that after coloring the scene seems natural at the same time it should be efficient in identifying the human presence. So the same method cannot be applied to both night vision infrared image and X-ray image. So two very different methods are proposed in the project for pseudo coloring of both different types of images.
1.4 Content Of The Report
The report starts with the basic introduction of the pseudo coloring. Chapter 1 discusses the basic need of applying pseudo color to gray scale image and its application in night vision camera and X-ray images. Chapter 2 discusses the various types of coloring techniques that are used in different fields. Then proceeding chapter discusses the pseudo coloring of night vision images using image fusion and how to apply the natural color to NIR images. The next chapter discusses the pseudo coloring of X ray images.
Chapter 2
Classification Of Pseudo Coloring Technique
Based on available literature, pseudo coloring techniques can generally be grouped into five main categories:
1) Spectrum-based maps, where color scales are designed by having the hue sequence
range from violet, via indigo, blue, green, yellow, and orange, to red, following the color order of the visible spectrum. Since the human visual system has different sensitivities to different wavelengths, researchers such as Clarke and Leonard indicated that spectrum-based color scales were not perceived to possess a color order that corresponds to the natural order of grayscale in the image.
2) Naturally ordered maps, where a number of researchers attempted to define a
particular path traversing the RGB color space under certain predefined constraints. The heated-object scale is achieved by bringing the RGB intensities up in the order of red, green, and blue, and limiting the path to 60◦ clockwise from the red axis. This selection is based on the claim that natural color scales seem to be produced when the intensities of the three primary colors rise monotonically with the same order of magnitude throughout the entire scale. To construct a natural color scale, Lehmann et al defined a spiral-like scale in the RGB model, keeping the original brightness progression of grayscale images. Specifically speaking, their color scale follows a spiral-like path along the diagonal of the RGB model. The authors formularized such a scale to allow the determination of the resulting number of colors, and demonstrated their scale’s effectiveness by applying it to medical X-ray images. They also pointed out that better results could be obtained if other color models were used. The use of HSI was suggested as future work based on the fact that the lightness component is directly represented by one of the axes in HSI.
3) Uniformly varying maps, where several studies focused on constructing a uniform
color scale where adjacent colors are equally spaced in terms of just noticeable differences (JNDs) and maintain a natural order along this color scale. Levkowitz and Herman’s research provided a scale with the maximum uniform resolution. The authors were hoping that their optimal color scale outperforms the grayscale, but evaluation results did not confirm that, at least for their particular application. They presented several reasons that might have caused
the unexpected results. One particular reason was that they used the CIELUV model to adjust the final colors which might not have been appropriate to model the perceived uniformity. Another reason was that the perceived change in color due to its surrounding was not taken into consideration. Shi et al designed a uniform color scale by visually setting up its color series to run from inherently dark colors to inherently light colors; i.e., from black through blue, magenta, red, yellow to white, then further adjusting the colors to make them equally spaced. The color scales were evaluated by comparing them to the grayscale. The authors indicated that the contrast sensitivity has been improved after applying their uniform scale, but they failed to demonstrate any significant outcome.
4) Function-based maps, where researchers utilized common mathematical functions
such as the sine function to construct desired mappings or color scales. Gonzalez and Woods described an approach where three independent mathematical transforms were performed on the gray level data, and the three output images fed into the R, G, and B channels to produce a specific color mapping.
5) Refernce Natural Images Database Technique
There are various approaches available in literature, which are broadly divided into three different categories: statistical approach, region-based approach, pattern recognition based approach. However from the recent literature, we found that combinations of these approaches are used to improve the quality of resultant image. Yang et al. in have proposed region-based approach for color transfer in night vision image sequences. The method uses support vector machines for region recognition among a set of natural color database. In case of specific application of color transfer in night vision FLIR (Forward Looking Infrared) images based on texture pattern recognition for color transfer images, Sun and Zhao proposed a method for automatic natural color mapping for FLIR. A local-coloring method utilizing image analysis and fusion was introduced by Zheng and Essock in, which render the NV image segment-by-segment by taking advantage of image segmentation, pattern recognition, histogram matching and image fusion. Recently, A. Toet proposed a fast color mapping method in which the mapping optimizes the match between the multi-band image and the reference natural image, and yields a night vision image with a natural daytime color appearance. This lookup-table based mapping procedure is simple and provides object color constancy. However the scene matching between source and target is performed manually
which may yield less naturalistic results for images containing regions that differ significantly in their content. Gang and Huang presented a multi-scale color image fusion using contourlet transform and expectation maximization (EM) where the color transfer is implemented in YUV color space. A new linear color space ICbCr was proposed by Xu and Li in especially for multiband night vision imagery to transfer the color distribution of the target image to the source NV images.
Chapter 3
Pseudo Coloring Of Night Vision Images
In the color transfer method color based clustering is applied on color map in LAB color space and cluster recognition is performed based on color similarity metric. The block diagram of the proposed color transfer method is shown in Figure 1. As shown in Figure 1, false colored nightvision image is obtained by assigning linear combinations of visible and IR source images to the RGB channels. The false colored NV image is enhanced by decorrelation stretch and contrast stretch. A colormap of the enhanced false colored
Fig. 1 Block Diagram Of Color Transfer Method
NV image is obtained. Color based clustering is performed on the colormap in the LAB color space using hill climbing algorithm. The association between each cluster and a natural color image in the target color look-up table is carried out automatically utilizing a nearest neighbour criteria based on a color similarity metric. The color components in each index within a cluster of the colormap are modified by statistics matching with the corresponding natural color image. The natural colored nightvision image is produced from the new colormap. Finally, the natural colored NV image is transformed to HSV color space and the enhanced gray image is substituted in the “value (V)” component to generate the final output of the proposed method. The subsequent subsections of this section describe the detailed explanation of the major steps of method [3].
3.1 False Color Fusion and Color Enhancement
The false color fused RGB image can be represented by the following equations: R(m; n) = (Vis(m; n) + IR(m; n)) G(m; n) = IR(m; n) B(m; n) = Vis(m; n) - IR(m; n) The false color fused image so formed has intensity variations similar to visible and IR source images. In order to achieve better seperation in color based clustering we perform decorrelation stretch for color enhancement and linear contrast stretch for intensity enhancement. Decorrelation stretch as described in increases color seperation across highly correlated channels while keeping the band variance same. Decorrelation stretch makes many features easier to observe which were not clearly visible in the original image.
3.2 Color based Clustering using Hill Climbing
Color based clustering is performed on the LAB colormap using the hill climbing algorithm. A color based image segmentation method using hill climbing algorithm proposed by Ohashi et al. is utilized for colormap clustering in the proposed method. The number of clusters required for proper classification of colormap are automatically determined by the hill climbing algorithm. A three-dimensional histogram is computed in the LAB color space. The hill climbing algorithm can be seen as a search window being run across the space of the 3D LAB histogram to find the largest bin within that window. Each bin in the color histogram has 3d - 1 = 26 neighbours where d = 3 is the number of dimensions of the feature space. The number of peaks found indicate the value of K and the value of those bins form the initial
seeds for the K-means segmentation. Thus, local maxima are found for clusters in the 3D color histogram of the colormap in the LAB color space. The entries of the colormap are then associated with the detected local maxima to generate several coherent clusters in the LAB colormap.
3.3 Cluster Recognition
The target color look-up table is created as follows. Each image from the natural color target image database is firsth smoothed by separately applying low pass filter in each RGB channels. This operation reduces the number of colors and enables the extraction of dominant colors from the natural image. The smoothed image is then transformed into LAB color space and first order statistics, mean µ and standard deviation , are computed for each band. A nearest neighbour criteria is used for automatic association of a cluster of colormap with a unique natural color image in the target color look-up table. The similarity metric
Where
Is used for the measure of the distance between Kth cluster and nth image.
3.4 Color Transfer
Color transfer is performed cluster-by-cluster by the standard statistics matching method proposed by Toet. Each index in LAB colormap is first checked as to which cluster does it belong and then the statistics of natural color target image associated with that cluster is transferred to the LAB values of the index in colormap. Thus a new LAB colormap is obtained. The equations for color transfer on each index in the colormap are defined as follows:
Kth cluster is associated with nth natural image. deviation of kth cluster in source images. deviation of nth target image.
and and
indicates the mean and standard indicates the mean and standard
Fig. 2 Results of Night Vision Color Image
Chapter 4
Pseudo coloring on X-ray images of Medical and Luggage scanning @ Airport
In this project we mainly focus on two major areas: 1. Medical Pseudo coloring is done to enhance the visibility of the fracture or any disease which is not clearly visible by naked eyes in x-ray image. 2. Luggage Scanning @ Air-port Pseudo coloring is done to increase the detect ability of the low density weapons which are not clearly detectable in simple x-ray scanning.
4.1 Simple Process of Pseudo Coloring
To develop enhanced color image from x-ray image, processing is done on x-ray image. A simple block diagram of that process is shown in figure 3. Few steps are followed before coloring to increase the visibility.
X-Ray Image
Enhanced Color Image
Contrast Stretch
Color Conversion
Salt & Pepper Noise Removal
Fig. 3 Simple Block Diagram of Pseudo coloring
A. X-ray Image: It is the simple black and white image taken by x-ray camera. Each pixel of the image has information i.e. intensity of which value varies from 0 to 255. Processing is done on this image. B. Contrast stretch This process increases the difference between maximum and minimum intensity. In most of the methods simple linear stretch is applied. C. Salt & Pepper Noise Removal After contrast stretching, Salt and pepper noise is removed. It is done by applying various filters. In most of the cases median filter is applied. At the end of this block we get enhanced gray image with more information regarding weapons in case of luggage scanning or fractures in case of medical images. D. Color Conversion It is the process of converting enhanced gray scale image into color image. Various methods for color conversion are available [1][2][5]. We have studied 4 basic methods: i. ii. iii. iv. Hot Jet HSI Rainbow
All the methods are further explained in detail.
4.2 Various Coloring Techniques
1. HOT It is RGB based linear color map. “Hot” changes smoothly from black, through shades of red, orange, and yellow to white as shown in fig. 4.
Fig.4 The HOT color Scale In fig. 3 RGB values versus gray value is shown for HOT coloring. That plot shows the variation of R,G and B with respect to I.
Fig. 5 R,G and B v/s I for HOT coloring
2. JET It is RGB based linear color map. It ranges from Blue to Red, passing through Cyan, Yellow and Orange as shown in fig.4. In fig. 5 RGB values versus gray value is shown for JET coloring. This color map can be obtained by converting this plot into simple mathematical formulas.
Fig. 6 The JET color scale
Fig. 7 R,G and B v/s I for JET coloring
3. HSI It is Histogram based non-linear color mapping. The colors of the various components in the scene are assigned based on the values of the raw image. Pixel ranges are selected from the data’s histogram and automatically given certain colors. For example, four gray-level regions were created, the chances of the low density threat being present would be greatest in the first two regions. Blue will be used as background and other easy-to-remember basic colors like red and green are applied to the other pixels in each bin. The output image would have four hues, which vary as a function of the gray intensity value of each pixel.
Fig. 8 H,S and I v/s gray for HSI based coloring
Fig. 9 HSI color scale based on concave part(solid part of fig. 8)
Fig. 10 HSI color scale based on convex part(solid part of fig. 8)
4. Rainbow It can be considered as a special case of Sine/Cosine transform. 3 color transfer functions are used for rainbow map. All the functions are periodic in the sense that they get peak in particular color interval. Rainbow based color map and RGB’s relation with I are shown in following figures.
Fig. 11 The Rainbow color scale
Fig. 12 R,G and B v/s I for Rainbow color map
Fig. 13 Simulation Results of various Coloring Methods
4.3 Proposed Method based on Look Up Table designed from Warm Color Scale
This method is specially designed for pseudo coloring in x-ray images for weapon detection and medical. In fig. 14 a detailed block diagram is shown.
X-Ray Image Warm Color Map Adaptive Histogram Equalization
Look Up Table Contrast Stretch Using Intensity Adjust
Noise Removal Using 2D-Median Filter
Enhanced Gray Scale Image
Color Conversion Using Look Up Table
Enhanced Color Image
Fig. 14 Detailed Block Diagram of Proposed Method based on Look Up Table designed from Warm Color Scale
Initially Look Up Table is created based on WARM color scale. Warm color scale is explained in [1][2][5]. A simple x-ray image is passed through various blocks to enhance the information regarding fractures/cracks or weapons in case of medical or weapon detection @ airport respectively. Pseudo coloring on this enhanced gray image is done based on the Look Up Table prepared earlier as per warm color scale.
Explanation of the above block diagram: [1]. X-Ray Image It is the simple black and white image taken by x-ray camera. It may be the image of any body part taken for medical use or may be image of weapon detection @ air-port. Each pixel of the image has information i.e. intensity of which value varies from 0 to 255. Processing is done on this image. [2]. Adaptive Histogram Equalization It is the process of enhancing the contrast of images by transforming the values in the intensity image I. Unlike simple histogram equalization it operates on small data regions (tiles), rather than the entire image. Each tile's contrast is enhanced, so that the histogram of the output region approximately matches the specified histogram. The neighboring tiles are then combined using bilinear interpolation in order to eliminate artificially induced boundaries. The contrast, especially in homogeneous areas, can be limited in order to avoid amplifying the noise which might be present in the image. In this proposed method tile size is taken as 16 x 16. [3]. Contrast Stretch using Intensity Adjust By this block linear intensity contrast stretch is applied. It increases the difference between maximum and minimum intensity. MATLAB inbuilt function imadjust is used for this function. [4]. Noise Removal Using 2D-Median Filter. At the end of the contrast stretching we get the image which may have few unwanted noise dots. It should be removed for false detection. This can be removed by filter. In this proposed method 2 dimensional median filter is used. It is the filter which takes the median of given square block. Here we have taken 3 x 3 block.
[5]. Enhanced Gray Scale Image This is the final gray scale image i.e. enhanced image having more information than original one. Now the coloring is done on this image. [6]. Warm color scale This is non-liner color scale. It varies from Dark Blue to Light Yellow through Magenta and Orange as shown in fig. 15. The distances of adjacent colors on this scale are perceivably equal. A 256-step scale as seen in Fig. 11 was developed. For any intensity I(i) and I(i+1) the correspondence (R, G, B)(i) < (R, G, B)(i+1). This law is followed throughout the color scale. Respective relation of R,G and B with I is shown in fig. 16.
Fig. 15 The WARM Color Scale
Fig.16 R,G and B v/s I for WARM coloring
[7]. Look Up Table It is a simple table which gives the corresponding R, G and B values for given intensity I. It is designed based on the Warm Color Scale which is already explained earlier.
[8]. Color Conversion using Look Up Table Color Conversion i.e. I to (R,G,B) is done based on the look up table created earlier. The coloring is done on enhanced gray scale image. It is just a simple assignment of (R,G,B) as per the value of intensity of that pixel.
[9]. Enhanced Color Image This is the final color image having more, clear information regarding fracture or cracks in medical images or weapons in luggage scanning @ airport.
Fig. 17 Simulation Results of Proposed method based on Look up table created as per warm color scale
Chapter 5
Evaluation Parameters for Coloring
The quality assessment of different image fusion schemes for X-ray images is traditionally carried out by subjective evaluations. The subjective evaluation is influenced by individual human perception. In recent literature [4] objective evaluation parameters are proposed. Here we have considered non reference based evaluation parameters; entropy and colorfulness metric. Entropy is used to measure the information content of an image. The entropy of a grayscale image is:
where G is the number of gray levels in the image’s histogram (which can be 255 for a typical 8-bit image) and p(i) is the normalized frequency of occurrence for each gray level, i.e., the histogram of the image. The entropy of colored image is computed for each band in RGB color space and average of the entropy of the three bands is considered for evaluation. Colorfulness metric is an efficient metric for calculating colorfulness of images and it is described in [4]. Larger the color variations in the image, higher is the colorfulness metric. The proposed algorithm is compared with the standard statistics matching method proposed by Toet. The simulation results of proposed algorithm are shown in Figure 17 and the simulation results of other coloring techniques are shown in figure 13. It is observed that proposed method provides more informative appearance compared to other method [1] and [5] as shown in Figure 13 and Figure 17. Hence the cracks in medical image and weapons in luggage scanning the scene can be easily distinguished. Their comparison with other techniques using parameters like entropy and colorfullness metric are shown in table 1.
ENTROPY
NO
1 2 3 4 5 6
HOT
0.9734 0.6764 0.7389 0.7643 0.6358 1.5171
JET
3.607 3.5149 3.6231 3.8605 3.5244 3.6661
HIS
5.4859 5.5525 5.5841 5.5473 5.7071 5.4594
RAINBOW PROPOSED
6.4439 6.4578 6.5618 6.443 6.9876 5.9877 5.772 5.5401 5.5839 5.5479 5.6924 5.7758
COLORFULLNESS METRIC
HOT
1 2 3 4 5 6 165.82 159.18 171.64 185.31 156.77 210.2
JET
0.8937 0.8947 0.9241 0.9396 0.811 1.073
HIS
0.989 0.9973 0.8892 0.9736 0.7456 0.8602
RAINBOW PROPOSED
0.3512 0.3512 0.3512 0.3512 0.3512 0.3512 76.3414 73.7114 73.7709 81.549 78.3135 87.2228
Table 1. Entropy and Colorfullness Mertic Comparision
CONCLUSION
Thus in this report we presented efficient method for pseudo coloring of grayscale images. Here we discussed the pseudo coloring of two different types of images. The first is Night vision infrared image and second is X ray images. Night vision images are used in army and nevy applications while the X ray images are used at air ports as well as in medical. Thus we can see that it is very crucial that the image should be visibly best as possible. So that personnel can extract information as much as possible. So we apply color to those images in this project. We can see that the colored images are far better to analyze than the gray images. Moreover same method cannot be applied to two different type of the field. So we used two very different methods to for night vision images and X ray images. For night vision images we try to apply natural color to images so that soldiers can relate the image to the surround environment and looks familiar. So we prepared the database using natural reference image and applied color transfer using that. In x ray we used the applied the standard color scales defined in the literature and applied. We also developed a color look up table for coloring and we observe that in this method we obtained the best result. In future we still find the another method for colorization. The pre-processing on the grayscale images is very crucial and more enhancing algorithms can be applied to obtain still better results.
References
[1]. Andreas Koschan and Mongi Abidi, "Digital Color Image Processing," A John Willy & Sons, INC., Publication, Hoboken, New Jersey. [2]. Rafael C. Gonzalez and Rechard E. Woods, “Digital Image Processing” Prentice Hall, New Jersey. [3]. Tanish Zaveri, Mukesh Zaveri, Ishit Makwana and Harshit Mehta,“ An Optimized Region-based Color Transfer Method for Night Vision Application”. [4]. [5]. Toet. Natural color mapping for multiband nightvision imagenary. Information Fusion, vol. 4(3), pp. 155-166, 2003. Besma R. Abidi, Senior Member, IEEE, Yue Zheng, Andrei V. Gribok, and Mo ngi A. Abidi, Member, IEEE. “Improving Weapon Detection in Single Energy X-Ray Images Through Pseudocoloring” Ieee transactions on systems, man, and cybernetics—part c: applications and Reviews, vol. 36, no. 6, pp. 784-796, November 2006.
