Analysis of Contrast Enhancement Techniques For Underwater Image

Analysis of Contrast Enhancement Techniques For Underwater Image Balvant Singh, Ravi Shankar Mishra, Puran Gour Abstract Image enhancement is a process of improving the quality of image by improving its feature. The underwater image suffers from low contrast and resolution due to poor visibility conditions, hence an object identification become typical task. In this paper comparative analysis of various contrast enhancement techniques for such underwater images is presented. The performance of contrast limited adaptive histogram equalization method is compared with contrast stretching, and histogram equalization method. For comparing the performance, mean square error and SNR used as parameters. The method is tested on various type of underwater image environment. Index Terms Image enhancement, contrast stretching, histogram equalization, Contrast limited Adaptive Histogram Equalization. I. INTRODUCTION Underwater image enhancement techniques provide a way to improving the object identification in underwater environment. There is lot of research started for the improvement of image quality, but limited work has been done in the area of underwater images, because in underwater environment image get blurred due to poor visibility conditions and effects like absorption of light, reflection of light, bending of light, denser medium (800 times denser than air), and scattering of light etc. These are the important factor which causes the degradation of underwater images. The researchers have reviewed several techniques related to images enhancement viz contrast stretching histogram equalization contrast limited adaptive histogram equalization (CLAHE). Here we are going to compare three enhancement techniques on the basis of SNR and mean square error as parameter. The remainsng of the paper is organized as follows: Next Section describe the various type of image enhancement method and algorithm, section II describe the contrast stretching techniques, in section III details histogram equalization with flow chart, section IV describe CLAHE method and its flow chart. Section V details result in comparative form. Section VI give conclusion and future work. 190 II. IMAGE ENHANCEMENT Image enhancement is a process of improving the quality of image by improving its features. Here we have given three method of image enhancement for underwater images (i) Contrast stretching (ii) Histogram equalization and Contrast limited adaptive histogram equalization. A number of contrast measures were proposed for complex images as underwater images [1,2,3,4]. A local contrast measure is proposed in this paper for enhancement. Contrast is stretched between the limit of lower threshold and upper threshold. It is an intensity based contrast enhancement method as I o (x,y) = f(i(x,y)), where the original image is I(x,y), the output image is I o (x,y) after contrast enhancement, and f is the transformation function. In the next stage histogram equalization is performe. Histogram is defined as the statistic probability distribution of each gray level in a digital image. Histogram equalization (HE) is one of the well-known methods for enhancing the contrast of given images, making the result image have a uniform distribution of the gray levels. It flattens and stretches the dynamic range of the image s histogram and results in overall contrast improvement. HE has been widely applied when the image needs enhancement however, it may significantly change the brightness of an input image and cause problem in some applications where brightness preservation is necessary. Finally in this paper, a novel CLAHE enhancement method is proposed which can yield the optimal equalization and also limit the contrast of the image. The suggested method is very useful for the video image broadcasting. where the brightness requirement is high such as in geographical channels. III. CONTRAST STRETCHING In local contrast measure is proposed in this project for enhancement. Contrast is stretched between the limit of lower threshold and upper threshold. It is an intensity based contrast enhancement method as Io (x,y ) = f ( I (x,y ) ), where the original image is I (x,y ),the output image is I o (x,y ) after contrast enhancement, and f is the transformation function

The contrast stretching is a method to make brighter portion more brighter and darker portion more darker. The transformation function T (r) is given as { ( ) ( ) Original grey level r Fig. 1 Contrast transformation function The transformation function is given here, Where l, m, n are the Slopes of the three regions shown in Fig.1. It is clear that l & n are less than one. The S is the modified gray levels and r is the original gray levels. Where a and b are the limit of lower and upper threshold. The identity transformation is shown by doted line. The slope of blue lines is taken 0.5 and the slope of the red line is taken as 1 or greater than 1. so making the brighter portion more brighter and darker portion more darker. The transformation that we are looking for must satisfy the following two conditions. (a) T (r ) must be single valued and monotonically increasing in the internal 0 r 1 (b) 0 T (r) 1 for 0 r 1 i.e. 0 s 1 for 0 r 1 The requirement in (a) that T(r) be single valued is needed to guarantee that the inverse transformation will exist, and the monotonic condition preserves the increasing order from black to white in the output image. A transformation function that is not monotonically increasing could result in at least a section of the intensity range being inverted, thus producing some inverted gray levels in the output image. The condition (b) guarantees that the output gray levels will be in the same range as the input levels. The inverse transformation from s back to r is denoted as. ( ) (2) For example 2 transformation functions are shown in Fig 2 III. HISTOGRAM EQUALIZATION There are many applications, wherein we need a flat histogram. This cannot be achieved by histogram stretching. so we use Histogram Equalization. A perfect image has equal number of pixels in all its grey levels. Hence to get a perfect image, our objective is not only to spread the dynamic range, but also to have has equal number of pixels in all its grey levels.this technique is known as Histogram Equalization (HE). Histogram equalization maps grey levels r of an image into grey levels s of an image in such a way that grey levels s are uniform. This expands the range of grey levels (contrast ) that are near the histogram maxima, and compresses the range of grey levels that are near the histogram minima. For most images, the contrast is usually expanded for most pixels, improving many image features. We now search for the transformation that would transform a original histogram to a flat histogramwe know that s =T( r ), so what is T which produces equal value gray levels. Fig.2 Transformation functions for Histogram Equalization The gray levels in an image may be viewed as random variables in the interval [0,1]. One of the most fundamental descriptors of a random variables is its probability density function ( PDF). So the grey levels for continuous variables can be characterized by their probability density functions p(r) and p(s). The probability density of the transformed grey level is shown in eq.2 is, dr P ( s) P( r) ds We now need to find a transformation, which would give us a flat histogram. Let us consider the Cumulative Density Function (CDF). Cumulative density function is obtained by simply adding up all the Probability density functions (PDF). 191

r ( ) = 0 (4) dr; 0 r 1 Since probability density functions are always positive, and recalling that the integral of a function is the area under the function so the transformation function iis the single valued and monotonically increasing, and satisfy the condition (a). similarly the integral of a probability density function for variables in the range [0, 1] also is in the range [0,1 so condition (b)is also satisfied.. Differentiating transformation function of eq 4 with respect to r we get ds P(r) (5) dr using eq 5 & 3 we get P(s) = [1] = 1 ; 0 s 1 P (s ) 1 1 s Fig. 3 a) Cumulative distribution function Because p(s) is a probability density function, it follows that it must be zero outside the interval [ 0, 1] in this case because its integral over all values of s must equal to 1.so the p(s) is a uniform probability density function. lets se what we get Fig 3 b) Transformation for the histogram equalization Using histogram equalization can be a good approach when automatic enhancement is desired, although there are still situations where basing image enhancement on a uniform histogram may not be the best approach. In these situations, histogram equalization effects may be too severe. So other histogram techniques may need to be used, such as adaptive histogram equalization IV. CONTRAST LIMITED ADAPTIVE HISTOGRAM EQUALIZATION (CLAHE) CLAHE was originally developed for medical imaging and has proven to be successful for enhancement of lowcontrast images such as portal films. The CLAHE algorithm partitions the images into contextual regions and applies the histogram equalization to each one. This evens out the distribution of used grey values and thus makes hidden features of the image more visible. The full grey spectrum is used to express the image. Contrast Limited Adaptive Histogram Equalization, (CLAHE) is an improved version of AHE, or Adaptive Histogram Equalization. Both overcome the limitations of standard histogram equalization. A variety of adaptive contrast limited histogram equalization techniques (CLAHE) are provided. Sharp field edges can be maintained by selective enhancement within the field boundaries. Selective enhancement is accomplished by first detecting the field edge in a portal image and then only processing those regions of the image that lie inside the field edge. Noise can be reduced while maintaining the high spatial frequency content of the image by applying a combination of CLAHE, median filtration and edge sharpening. A variation of the contrast limited technique called adaptive histogram clip (AHC) can also be applied. AHC automatically adjusts clipping level and moderates over enhancement of background regions of portal images. The expression of modified gray levels for standard CLAHE method with Uniform Distribution can be given as [ ] ( ) (8) For discrete values we deal with probability s and summations instead of probability density functions and integrals. nk k = 0,1,2,3,..L-1 (6) n And the discrete version of the transformation function is given by, S k = T( r k ) = P( r j ) s k = nk k = 0,1,2,3,..L-1 (7) n 192 Where Maximum pixel value Minimum pixel value g is the computed pixel value ( )=CPD (Cumulative probability distribution) For exponential distribution gray level can be adapted as Where is the clip parameter? ( ) [ ( ) ] (9)

a) Original See fish image b) Contrast stretched image c) Histogram Equalized image d) CLAHE Image e) Original Fish plant image f) Contrast stretched image g) Histogram Equalized image h) CLAHE Image i) Original See plant image j) Contrast stretched image k) Histogram Equalized image l) CLAHE Image m) Original See object image n) Contrast stretched image o) Histogram Equalized imaged) CLAHE Image Fig. 5 Comparison of various contrast enhancement methods In Fig. 5 results of the various enhancement methods are presented. For contrast stretching upper and lower threshold values are 200 and 50 respectively. The slope of stretching curve is 0.5, 1.5 and 0.5. For CLAHE method image is divided into tiles of size is 8x8, and clip limit is taken 0.02 and distribution is uniform. It is clear that the CLAHE method not only gives better equalization but also improves the contrast of image. This can be clearly seen from the histograms of the results in a) Original Histogram b) Histogram of Contrast stretching c) Histogram of Equalization d) Histogram of CLAHE method Fig. 6 Comparison of Histograms for See plant image The comparison of MSE and SNR are given in Table 1 and 2. It is clear that CLAHE method improves SNR. Table 1 Comparison of MSE S. Images Contrast Histogram CLAHE No Stretching Equalization 1 See Fish 32.4317 64.72 34.771 2 See Plant 17.295 43.462 26.967 3 Fish plant 16.4221 47.68 13.378 193

Table 1 Comparison of SNR S. Images Contrast Histogram CLAHE No Stretching Equalization 1 See Fish 5.95 2.996 5.558 2 See Plant 6.895 2.7643 4.4257 3 Fish plant 7.842 2.722 9.627 V. CONCLUSION An comparison of enhancement method is presented in the paper. It is found that CLAHE method not only improve the contrast but also equalizes the image histogram efficiently. It is observed that SNR is improved with CLAHE method in comparison with HE. The enhancement methods effectively improves the visibility of underwater images. REFERENCES [1] A. Daskalakis, D. Cavouras, P. Bougioukos, S Kostopoulos, and George Nikiforidis, An efficient CLAHE based, spot adaptive, image segmentation technique for improving microarray genes quantification, 2nd Int.l Conference on experiments Process / /System Modelling/Simulation & Optimization 4-7 July, 2007 [2] Sos S. Agaian,,, Blair Silver, and Karen A. Panetta,, Transform Coefficient Histogram-Based Image Enhancement Algorithms Using Contrast Entropy, IEEE Transaction on image processing, VOL. 16, NO. 3, MARCH 2007. [3] R. Dale-Jones and T. Tjahjadi, A study and modification of the local histogram equalization algorithm, Pattern Recognition, vol. 26, no. 9, pp. 1373 1381, 2007. [4] Tristan J. Lambert, Digital Enhancement Techniques for Underwater Video Image Sequences, submitted at School of Computing In Nov.2005, [5]. K. Zuiderveld, "Contrast Limited Adaptive Histogram Equalization", Graphics Gems IV, pp. 474-485 [6] B. Silver, S. S. Agaian, and K. A. Panetta, Logarithmic transform coefficient histogram matching with spatial equalization, presented at the SPIE Defense and Security Symp., Mar. 2005. [7] B. Silver, S. S. Agaian, and K. A. Panetta, Contrast entropy based image enhancement and logarithmic transform coefficient histrogram shifting, presented at the IEEE ICASSP, Mar. 2005. [8] Etta D, Pisano, S. Zong, R. E Jhonston Contrast limited adaptive histogram equalization image processing to improve the detection of simulated speculation in Dense Monograms, Journal of Digital Imaging, vol. 11, No. 4, pp 193-200, Nov. 1998 [9]. R. Gonzalez, R. Woods, Digital Image Processing, 2nd Edition, Prentice Hall, 2002. [10] Gonzalez, RC, Woods, RE & Eddins, SL 2004, Digital Image processing using MATLAB, Pearson Prentice Hall. [11] Kashif Iqbal, Rosalina Abdul Salam, Azam Osman and Abdullah Zawawi Talib Underwater Image Enhancement 194 Using an Integrated Colour" Model,IAENG International Journal of Computer Science, 34:2, IJCS_34_2_12, 2009. [12] Shilong Wang, Yuru Xu and Yongjie Pang. A Fast Underwater Optical Image Segmentation Algorithm Based on a Histogram Weighted Fuzzy C-means Improved by PSO, journal of Machine science, No.10 pp 70-75, April 2011. [13] Schechner, Y & Karpel, N, 'Clear Underwater Vision', Computer Vision & Pattern Recognition, vol. 1, pp. 536-43, 2008 [14] Petit. F, Capelle Laize, Carre, P., Underwater image enhancement by attenuation inversionwith quaternions, IEEE International Conference on Acoustics, Speech and Signal Processing, ICASSP 2009, 26.may 2009. Author s Profile: 1. Balwant Singh BE in Electronics & Communication, currently pursing M. Tech in Digital Comm. From NRI Bhopal. 2. Dr. Ravi Shanker Mishra, Ph.D from MANIT Bhopal, and Currently working as Prof. in NRI College 3. Prof. Pooran Gour, BE, M.Tech, and pursuing Ph.D., & Currently working as Prof. in NRI College