A Survey of Image Enhancement Techniques

Transcription

1 A Survey of Image Enhancement Techniques Sandeep Singh, Sandeep Sharma GNDU, Amritsar ABSTRACT This paper has focused on the different image enhancement techniques. Image enhancement has found to be one of the most important vision applications because it has ability to enhance the visibility of images. It enhances the perceivability of poor pictures. Distinctive procedures have been proposed so far for improving the quality of the digital images. To enhance picture quality image enhancement can specifically improve and limit some data presented in the input picture. It is a kind of vision system which reductions picture commotion, kill antiquities, and keep up the informative parts. Its object is to open up certain picture characteristics for investigation, conclusion and further use. The main objective of this paper is to discover the limitations of the existing image enhancement strategies. Keywords: Image enhancement, human visual perception, Visibility. 1. INTRODUCTION Image enhancement [12] is mainly increasing the observation of information in images for individual viewers and providing improved input for other image processing methods. The principal aim of image enhancement is to alter attributes of an image to make it more appropriate for a given job and a specific observer. During this process, one or more attributes of the image are altered. Figure 1 Before and After Enhancement The choice of attributes and the way they are altered are definite to a given job. Moreover, observer-definite factors, such as the human visual system and the observer s knowledge, will introduce a huge deal of non-objective into the option of image enhancement techniques. There are many methods that can improve a digital image without ruining it. 2. NON LINEAR ENHANCEMENT There are some problem in previous methods i.e. after enhancement the image details are degraded. So to eradicate this problem the luminance and contrast enhancement of V component of the input image has been described. The form of the nonlinear transfer function [1] for luminance enhancement is equal for every pixel. But the illumination all over the image has not been same, some regions may be dark and some may be bright. Thus the image locality has to be considered while enhancing the color image. In order to achieve this goal firstly the value component image in HSV image space has been divided into smaller overlapping block sand find the shape of the nonlinear transfer function for each pixel to enhance the luminance of the image. In contrast enhancement process, for each pixel the amount of enhancement has been calculated depending upon the center pixel itself and its surrounding pixel values. Fig. 2 shows the block diagram of the proposed method. Volume 2 Issue 5 May 2014 Page 1

2 In general, color images are represented in RGB color space. HSV space is closer to human perception in which the (H) refers to the spectral composition of color, saturation (S) defines the purity of colors and (V) refers the brightness of a color or just the luminance value of the color. The RGB values of an image are converted into HSV values. The luminance enhancement also known as process of dynamic range compression, which is the first step for enhancing the images taken under low or discontinuous illumination environment, has been applied to the significant component of the input image using specially designed non-linear transfer function. The V channel image has been subjected for luminance improvement. Assume that VLE be the transferred value by applying varying transfer function as follows: Figure 2 How non-linear image enhancement works (adapted from [1]) 3. HISTOGRAM SPECIFICATION Dynamic Histogram Specification [13] algorithms based on the Histogram Specification (HS) methods have been proposed to enhance the contrast without losing the genuine histogram characteristics. In order to keep genuine histogram features, the DHS extracts the differential information from the input histogram. On the other hand, it also applies extra parameters to control the overall processing, such as frame direct current and gain control value. Figure 3 (a) Original Histogram Figure 3 (b) Histogram Specification Volume 2 Issue 5 May 2014 Page 2

3 4. CRITICAL ANALYSIS OF VARIOUS TECHNIQUES The objective of the literature review is to find and explore the benefits of image enhancement algorithms and also to find the short comings in existing algorithms and techniques. The main goal of this literature review is to find the gaps in existing research and techniques and possible solutions to overcome these holes. 5. LITERATURE SURVEY KIM (1997) [1] has discussed that the intensity of scene can be altered following the histogram equalization, which is because of the flattening attribute of the histogram equalization. KIM (1997) [6] proposed histogram equalization referred to as mean preserving bi-histogram equalization to conquer the disadvantage of the histogram equalization. The essence of the proposed algorithm is to conserve the mean intensity of an image while the contrast is enhanced. Given method initially break an input image into two sub-images based on the mean of the input image. One of the sub-images is the set of specimens that are less than or equal to the mean whereas the other one is the set of specimens greater than the mean. Tae et al. (1998) [2] has discussed a block-overlapped histogram equalization system for improving the contrast of an image sequences using numerous applications. The conventional histogram-based contrast enhancement technique is limited in real time application due to a large computational and storage requirements and it also exhibit quality degradation caused by possible loss of infrequently distributed pixel intensities, which may result in terrible loss of vital information. Yueet et al. (2005) [3] has discussed a non-linear image enhancement method based on Gabor filters, which allows selective enhancement based on the contrast sensitivity function of the human visual system. The image enhancement of the given approach is especially appropriate for digital applications to enhance the perceive visual feature of the images due to numerous reasons, including interpolation. Saibabu et al. (2006) [4] has proposed an image enhancement algorithm for digital images captured under such tremendously non-uniform lighting conditions. The new technique constitutes three issues viz, adaptive intensity enhancement, contrast improvement and color restoration which were considered separately to make the algorithm more adaptable to the image characteristics. The adaptiveness of the transfer function, depending on the mean of each pixel s neighborhood makes the algorithm more flexible and easier to control. Nyamlkhagva et al, (2008) [5] has proposed a new method called Brightness Preserving Weight Clustering Histogram Equalization that can simultaneously preserve the brightness of the original image and enhance visualization of the original image. Given method assigns each nonzero bit of the original image s histogram to a separate cluster, and computes each cluster s weight. To reduce the number of clusters, three criteria are used (cluster weight, weight ratio and widths of two neighboring clusters) to merge pairs of neighboring clusters. The clusters obtain the equal partitions as the result image histogram. At last, transformation functions for each cluster s sub-histogram are calculated, and the sub-histogram s gray levels are mapped to the result image by the equivalent transformation functions. Fan et al. (2010) [6] has proposed a new method for image contrast enrichment which is especially suitable for multiple-peak images. The given method has been used to remove the two disadvantages of HE algorithm i.e. firstly the input image has been convolved by a Gaussian filter with optimum parameters. Then, the original histogram has been divided into various areas by the valley values of the image histogram. The given method out performs others on the aspects of simplicity and adaptability. The result demonstrates that the proposed algorithm has good performance in the areaof image enrichment. Due to its simplicity, it can be realized by simple hardware and consumer electronics. Md, Foisal et al. (2010) [7] has proposed a method of medical image enhancement based upon non-linear technique and the logarithmic transform coefficient histogram equalization. Logarithmic transform histogram matching uses the truth that the relation between stimulus and perception is logarithmic. A measure of improvement based on the transform has been used as a tool for evaluating the performance contrast measure with respect of the proposed enhancement technique. This method improves visual quality of images that contain dark shadows due to limited dynamic range of imaging like x-ray images. Kwok et al. (2010) [8] has proposed a strategy of local sector improvement by histogram equalization. In given strategy the image has been first divided into sectors and they are independently enhanced by histogram equalization, intermediate images are then provoked recursively by making use of this approach and a resultant image has been obtained by a weighted-sum aggregation on the basis of an intensity gradient measure. Local sectors with higher contrast dominate the others thus achieving overall global contrast improvement. An enhanced image is then produced where the intermediate images are repeatedly averaged using a weight-sum strategy. Cheng et al. (2012) [9] has discussed a novel approach for the detection of over- enhancement. Firstly the causes for generating over - enhancement has been analyzed in detail, then an objective and effective criterion has been presented. The experimental results demonstrate that the given approach can locate the over enhanced areas accurately and Volume 2 Issue 5 May 2014 Page 3

4 effectively, and provide a quantitative criterion to assess the over-improvement levels well. The given method will be useful for vigorously monitoring the quality of the improved image, and optimizing the parameter settings of the contrast improvement algorithms. Deepak et al. (2012) [10] has proposed a method for improving the color images based on non-linear transfer function and pixel neighborhood by conserving details. In given method, the image improvement has been applied only on the V (luminance value) component of the HSV color image and H and S component are kept unaffected to prevent the degradation of color balance between HSV components. The V channel has been enhanced in two steps. First the V component image has been divided into little overlapping chunks and for each pixel inside the chunk the luminance improvement is carried out using non-linear transfer function. Secondly, each pixel has been further improved for the adjustment of the image contrast depending upon the center pixel value and its neighborhood pixel values. Finally, original H and S component image and enhanced V component image are converted back to ROB image. Rajib et al. (2012) [11] has proposed a contrast enhancement technique using scaling of internal noise of a dark image in discrete cosine transform domain. The mechanism of improvement is attributed to noise-induced transition of discrete cosine transform coefficients from a poor state to an improved state. This transition is effected by the inner noise present due to lack of adequate illumination and can be modeled by a general bi-stable system exhibiting dynamic stochastic resonance. The given approach has adopted a limited adaptive processing and significantly improves the image contrast and color information while ascertaining good perceptual quality. When compared with the present improvement approaches such as adaptive histogram equalization, gamma correction etc. the given approach has shown extraordinary performance in terms of relative contrast enhancement, colorfulness and visual quality of improved image. 6. GAPS IN EARLIER WORK After conducting the literature survey I found that the basic limitation of the transform-based image enhancement introduces certain artefacts like:- a. They cannot at the same time improve all parts of the image very well and it is difficult to automate the image improvement procedure. b. The main problem of transfer-based image improvement is that, after enhancement the image detail are ruined. 7. POSSIBLE SOLUTIONS The non-linear enhancement algorithm uses HSV plane to enhance the image, as it is known H and S component need no modification for enhancement so alteration will be done on V (intensity) only. The one possible solution to overcome the discussed problem in section 5 will comprise two processes, i.e. adaptive intensity enhancement and contrast enhancement. Adaptive intensity enhancement utilizes a particularly designed nonlinear transfer function which is proficient of reducing the intensity of light regions and at the same time enhancing the intensity of gloomy regions. Contrast enhancement tunes the intensity of each pixels magnitude based on its nearby pixels. But it is also known that the enhancing the light or value component may not produce effective results in all the cases because change in brightness or light can only change the sharpness of the image. But it has no ability to modify the color strength which is also required in image enhancement. So to improve the nonlinear image enhancement technique one can integrate non-linear image enhancement with the histogram specification. Histogram specification has ability to adjust the colors and saturation of the image efficiently; so will produce better results. However as nonlinear image enhancement has produced accurate results for intensity, so we will apply histogram specification only for H and S. This will also ensure that the quality of the image will be improved. 8. CONCLUSION& FUTURE WORK The image enhancements techniques play a significant role in digital image processing. It is shown in this paper that the nonlinear image enhancement can be used to improve the quality of a blurred image by using the concept of the light source refinement. But in the most of techniques it has been found that the available technique does not provide better results in multiple light sources because no modification is done on the hue and saturation. As discussed earlier the image enhancement technique can be improved by modifying the hue and saturation. It will provide better results than the existing techniques. This work has focused on the limitation of the existing work and finding the possible solution for the same. In near future suitable tool will be used to implement the possible solution discussed in section 6. Also some more adjustment will be done by using the L*a*b color space instead of the HSV only. Volume 2 Issue 5 May 2014 Page 4

5 References [1] Kim. Yeong-Taeg. Contrast enhancement using brightness preserving bi-histogram equalization. Consumer Electronics, IEEE Transactions on 43, no. I (1997): 1-8. [2] Kim, Tae Keun, Joon Ki Paik. and Bong Soon Kang. Contrast enhancement system using spatially adaptive histogram equalization with temporal filtering. Consurner Electronics, IEEE Transactions on 44, no. I (1998): [3] Yang, Yue. andbaoxin Li. Non-linear image enhancement for digital TV applications using Gabor filters. In Multimedia and Expo, (2005). ICME (2005). IEEE International Conference on, pp. 4-pp. IEEE. (2005). [4] Angela. Saibahu, and K. VijayanAsari. An Adaptive and Non Linear Technique for Enhancement of Extremely High Contrast Images. In Applied Imagery and Pattern Recognition Workshop. (2006). AIPR (2006). 35th IEEE, pp IEEE. (2006). [5] Sengee, Nyarnlkhagva. and Heung Choi. Brightness preserving weight clusteringhistogram equalization. Consumer Electronics, IEEE Transactions on 54, no. 3 (2008): [6] Yang. Fan. and Jin Wu. An improved image contrast enhancement in multiple-peak images based on histogram equalization.t In Computer Design and Applications (ICCDA), (2010) International Conference on. vol. 1, pp. Vl IEEE, (2010). [7] Hossain, MdFoisal, Mohammad Reza Alsharif, and Katsumi Yamashita. Medical image enhancement based on nonlinear technique and logarithmic transform coefficient histogram matching. In Complex Medical Engineering (CME). (2010) IEEE/ICME International Conference on. pp IEEE. (2010). [8] Kwok. Ngai Ming. QuangPhuc Ha. Gu Fang. Ahrnad 13. Rad, and Dalong Wang. Color image contrast enhancement using a local equalization and weighted sum approach. In Automation Science and Engineering (CASE). (201 0) IEEE Conference on, pp IEEE. (2010). [9] Cheng, H. D., and Yingtao Zhang. Detecting of contrast over-enhancement: In Image Processing (ICIP), (2012) 19th IEEE International Conference on, pp IEEE, (2012). [10] Ghim ire, Deepak, and ioonwhoan Lee. Nonlinear transfer function-based local approach for color image enhancement. Consumer Electronics, IEEE Transactions on 57. no. 2(2011): [11] Jha. Rajib Kumar, RajlaxrniChouhan. Prabir Kumar Biswas, and KiyoharuAizawa. Internal noise-induced contrast enhancement of dark images. In Image Processing (ICIP), (2012) 19th IEEE International Conference on, pp IEEE. (2012). [12] Maini. Raman. and 1-limanshuAggarwal. A comprehensive review of image enhancement techniques. arxiv preprint arxiv: (2010). [13] Sun, Chi-Chia. Shanq-Jang Ruan. Mon-ChauShie, and Tun-Wen Pai. Dnamic contrast enhancement based on histogram specification. Consumer Electronics, IEEE Transactions on 51, no.4(2005): AUTHOR Sandeep Singh has done his B.Tech from Amritsar College of Engineering and Technology, Amritsar, Punjab, INDIA in Computer Science and Engineering with 66 percentage. Now he is pursuing M.Tech in Computer Science from GNDU, Amritsar, Punjab, INDIA. His area of interest is Digital Image Processing, Database Management, Computer Networks, Software Engineering, Computer Architecture. Sandeep Sharma has done his B.E in Computer Science and Engineering, M.E in Computer Science and Engineering, P.HD. He has published his papers in various National and Internatonal Journals. His area of interest is Digital Image Processing, Multistage Interconnection Network, Cloud Computing, Object Oriented Programming. His area of specialization is parallel processing. Now he is doing job as Associate Professor in Guru Nanak Dev University, Amritsar, Punjab, INDIA. Volume 2 Issue 5 May 2014 Page 5