Contrast Enhancement Based on BAT Algorithm

In this research paper, an algorithm is developed to produce better contrast enhancement which is inspired from nature and calculates the upper and lower limit for each sliding window. These upper and lower limits are used to calculate the local mean and global mean. It is based on contrast objective function; this is parabolic logarithmic threshold function. The results show that, the method is able to produce better contrast sensitivity and pleasing visuals as compared to older methods( histogram equalization, adaptive histogram equalization, fusion of pyramid and Gaussian , ant colony optimization method etc). The corpus of image consists of both gray and color images. Other than these other evaluation values like loudness, pulse rate, frequency show that this method provides better contrast ratio. KeywordsBat algorithm, Contrast Enhancement, Fusion Method INTRODUCTION Contrast enhancement is a commonly used operation in Image Processing. Contrast, is a property of image which makes the visual appearance better by adding color grading in the image. Contrast is calculated in terms of color and brightness of the object and other objects within same view of field. Contrast enhancement is the operation which is used in Medical images. It plays an important role for processing scientific images such as X-rays or satellite images. There are many methods proposed to enhance the contrast of an image. The most simplest and effective method for contrast enhancement is Histogram Equalization. The basic idea behind Histogram Equalization was to rescale the gray levels of image. Histogram Equalization removed the annoying artifacts and unwanted noise. The limitation of histogram equalization was that it applied on entire image whereas this limitation was overcome by Adaptive histogram Equalization which operated on small regions/blocks. The improved version of Adaptive histogram Equalization was Contrast Limited Adaptive Histogram Equalization which partitioned images into contextual regions and then histogram equalization was applied. This process evens out the distribution of gray pixel values and make out the hidden details visible to human. To improve contrast there was a Dualistic Sub Image Histogram Equalization in which the original image was decomposed into two equal sub images based on its gray level probability density function. On the basis of normalized law function they changed the sub histogram through weighting process and then equalized the weighted sub Histogram Enhancement of contrast using Weighted Threshold Histogram Equalization with Improved Switching Median Filter. The general idea adopted by WTHE was to modify the histogram and assigned weight and threshold to each pixel before equalization. Filtering played a vital role in signal processing and main goal of filtering was to enhance fine details of an image. The image was enhanced by WTHE then it passed through Improved Switching Filter, in improved switching median filtering it modified the corrupted signal without affecting uncorrupted signal and reduced the impulse noise created during enhancement. The performance of this method was evaluated by Absolute Mean Brightness Error (AMBE), Measure of Enhancement (EME), PSNR, and MSE. There were few bio inspired algorithms applied for contrast enhancement such as Ant Colony Optimization, Improving Ant Colony Optimization, Local and Global Contrast Algorithm, Firefly Algorithm. International Journal of Engineering Research and General Science Volume 3, Issue 4, Part-2, July-August, 2015 ISSN 2091-2730 182 www.ijergs.org PROPOSED WORK Bat Algorithm 1. Objective function f(x), x = (x1; :::; xd) T 2. Initialize the bat population xi for i:::::::::n vi= pulse distance/ time fi = fmin + [ fmax-fmin] * β where β is frequency adjustment constant. 3. Let Amin to Amax be the loudness range dependent upon frequency and emission rate as the emission rate slows when the loudness increases and vice versa. 4. Let λ be the wavelength at a fixed value For each iteration, xo be the starting search point. Get Position x, Loudness A, frequency f, Wavelength λ Generate Random solution. If new solution is better than old solution Terminate Else 5. Update position, loudness, frequency, emission rate. Generate Random number between 0 and 1. If (rand> pe) Select best solution Generate local solution among best solution If(rand< Ai) and f(x1) < f(x2)) Accept solution Increases pe( pulse emission rate) and reduce A( loudness) International Journal of Engineering Research and General Science Volume 3, Issue 4, Part-2, July-August, 2015 ISSN 2091-2730 183 www.ijergs.org End Table I Showing BAT Parameters and its values Sno BAT Parameters Values 1 Population Size (Bat Number) 20 2 Loudness 0.5 3 Minimum Frequency 0.1 4 Maximum Frequency 2 5 Pulse Rate 0.4 6 Number of Generation( Iterations) 1000 7 f(x), x = (x1; :::; xd) Objective Function [0,1] 8 Upper Bound 4 9 Lower Bound 1 10 Design of Experiment for Optimal parameters method 5 Table II Showing Research Parameters and its values Sno Research Parameters Values 1 Number Of Images 15 2 Size of Images 8KB-1.12MB 3 Contrast Upper value 1 4 Contrast Lower value 0.1 International Journal of Engineering Research and General Science Volume 3, Issue 4, Part-2, July-August, 2015 ISSN 2091-2730 184 www.ijergs.org FLOW CHART RESULTS In the previous work researchers enhanced the contrast of an image but with some limitations such as unwanted noise, artifacts, more computation time. In the current research work it is wise to enhance the contrast of an image by using nature inspired algorithms which gives the better results (a) (b) (c) Increase pulse rate and decrease loudness Find the Best Solution Initialize position and velocity Define Pulse frequency Initialize pulse rate and loudness Generate New Solution Update position, frequency, pulse emission rate Generate new solutions by random fly Set objective function f(x) International Journal of Engineering Research and General Science Volume 3, Issue 4, Part-2, July-August, 2015 ISSN 2091-2730 185 www.ijergs.org Fig1 : (a) Original Image (b) Apply fusion method of Pyramid and Gaussian (c) Proposed Method (a) (b) (c) Fig2 : (a) Original Image (b) Apply fusion method of Pyramid and Gaussian (c) Proposed Method (a) (b) (c) Fig3 : (a) Original Image (b) Apply fusion method of Pyramid and Gaussian (c) Proposed Method (a) (b) (c) Fig4 : (a) Original Image (b) Apply fusion method of Pyramid and Gaussian (c) Proposed Method Table III International Journal of Engineering Research and General Science Volume 3, Issue 4, Part-2, July-August, 2015 ISSN 2091-2730 186 www.ijergs.org Comparison with Existing Method Images Fusion of Gaussian and Pyramid Proposed Method Fig 1 84.5738 87.7512 Fig 2 97.8488 99.2395 Fig 3 66.050 73.6042 Fig 4 39.7182 43.1568 This table shows that the proposed method which is bat based contrast enhancements gives better results as compared to existing method which was based on fused based. CONCLUSION In summary, we can say that contrast enhancement methods must be able to produce images that are not just pleasing to human eyes and perception, but also must be able to produce more information within the content of image. There are many methods implemented till date, and most of the methods were based upon stretching the edges of objects. In such a way, difference in gray levels or color levels increases and the image does not lose its quality, or may introduce unwanted artifacts when image is reproduced after contrast enhancement algorithm applications. In this research work, we were able to do a better treatment in context of contrast, as it is apparent from the evaluation parameters. The results produced by proposed method shows better results as compared to the existing fusion based method. The existing method gives values in terms of grey level variance 40.5807, 63.9614, 66.050 of some images whereas proposed method gives 44.5555, 65.0935, 73.6042 and this shows the better contrast enhancement. FUTURE SCOPE In future, we suggest the proposed method may be extended for medical images that follow some standards like Health Level 7 (HL7), as these images need multi-resolution, multi-contrast view to satisfy at particular diagnostic process. REFERENCES: [1] Chao Zuo, Qian Chen, Xiubao Sui, and Jianle Ren, “Brightness Preserving Image Contrast Enhancement Using Spatially Weighted Histogram Equalization” volume11, No 1, January 2014 [2] Kesharee Singh Yaduwanshi, Nitin Mishra “ Contrast Enhancement of HDR images using linear transformation and kernel padding” volume5(2),2014 [3] Shyam Lal , Mahesh Chandra ,” Efficient Algorithm for Contrast Enhancement of Natural Images, vol 11, No1 , January 2014 International Journal of Engineering Research and General Science Volume 3, Issue 4, Part-2, July-August, 2015 ISSN 2091-2730 187 www.ijergs.org [4] Shih-Chia Huang, Fan-Chieh Cheng, and Yi-Sheng Chiu,” Efficient Contrast Enhancement Using Adaptive Gamma Correction with Weighing Distribution, volume 22, issue 3, march 2013 [5] Qing Wang, Rabab Ward ,” Fast Image/Video Contrast Enhancement based on WTHE” [6] Amiya Halder ,”Dynamic Contrast Enhancement Algorithm”, volume 74 no-12 july 2013 [7] Ritu Chauhan, Sarita Singh Bhadoria ,” An improved image contrast enhancement based on histogram equalization and brightness preserving weight clustering histogram equalization”, 2011 International Conference on Communication Systems and Network Technologies (CSNT), vol., no., pp.597,600, 3-5 June 2011 [8] Nirmal Singh, Maninder Kaur, K.V.P Singh , “Parameter Optimization Using PSO”, volume 2, issue 3 , 2013 [9] R.Sharmila, R.Uma ,” A new approach to image contrast enhancement using weighted threshold equalization with improved switching median filter”volume 7, issue 2 , 2014 [10] Pooja Mishra ,Mr.KhomLal Sinha,” Different Approaches of Image Enhancement” (2014)