International Journal of Computer Applications (0975 – 8887) Volume 122 – No.22, July 2015 13 Random Walker Segmentation based Contrast Enhancement of Dark Images with Canny Detection Harshit Khare Department of Computer Science & Engineering Oriental Institute of Science & Technology Bhopal, India Sanjay Sharma Asst. Prof., Department of Computer Science & Engineering Oriental Institute of Science & Technology Bhopal, India ABSTRACT Contrast enhancement is a technique which enables images to improve the contrast level of images. Contrast enhancement of images requires filtering of regions where contrast level is high or where noise level is more. The techniques such as non-dynamic based stochastic resonance are implemented but the technique provides less accuracy of contrast improvement. Hence an efficient technique is implemented here by segmented the low contrast region of the image and then filtering is performed on the segmented region using transformation. The proposed methodology greatly improves the contrast enhancement of the images. Keywords DWT, Bilateral Filtering, Gaussian Noise, Segmentation, Mosiac images. 1. INTRODUCTION Images play a vital role in various fields for the processing of various tasks such as in science and medicine and journalism as well as advertisement and design and education, entertainment. Hence the feature especially spatial features are necessary to extract such that it can be applied in various research fields. Usually, the raw data of a set of images is investigated to gain just around the corner into what is come about with the images and how they can be used to extract aspiration information. There are certain images that are taken in extreme lighting condition and sometimes contrast enhancement of these images are necessary to obtain. The enhancements of these images are necessary for the processing in various tasks [1]. The contrast enhancement of images is very necessary for the processing especially in computer vision. The enhancements are necessary for various research and field areas such as speech recognition [2]. Many images have very low dynamic range of the intensity values due to inadequate enlightenment, and so consequently need to be processed before being demonstrated. Many techniques for contrast enhancement that operate in spatial domain subsist in literature [3], [4]. In the process of Image Processing noise in the images play a vital role [5, 6]. 2. LITERATURE SURVEY In this paper [7], has proposed a new and effective technique for improvement of shadowy and short contrast images a nonlinear non-dynamic stochastic resonance based method. By taking care of a low contrast image as a sub threshold signal and adding random noise-enhanced signal processing is functional to get better its contrast tracked by hard- thresholding and averaging Random noise is added frequently to an image and is consecutively hard-threshold followed by taken as a whole averaging. By show a discrepancy the noise intensities, noise induced resonance is get hold of at exacting optimum noise intensity. Here author [7] unearth out the concert of the planned practice has been considered for four types of noise distributions - Gaussian, uniform, Poisson and gamma was investigation and best concert was examined for Gaussian model. Quantitative assessments of their presentations have been done in expressions of contrast enhancement factor, color enhancement and perceptual quality compute showed that it reaches higher performance metrics as evaluated to existing spatial domain methods. Comparison with other be presenting spatial domain techniques give you an idea about that the proposed method gives significant enhancement while determining good perceptual quality. In this paper [8] author has make use of a quantitative, strong calculation to estimate the spatial exposure of feature points in an image, intelligent to establish whether points are wide- ranging at manifold balance. When identical images for relevance’s such as mosaicking and homography assessment, the giving out of facial appearance across the overlie section affects the accuracy of the result show that SFOP commences extensively less aggregation than the other detectors tested and it is measured by Ripley’s K-function, to evaluate whether feature matches are clustered simultaneously or increase approximately the have common characteristics region. Based on this determine, an appraisal of a range of up to date feature detectors and then carried out using analysis of variance and a large image database was executed; the estimate method considered the imagery and the detector as the two self-determining variables have an effect on coverage, and consequence was evaluated using ANOVA. The results revealed that there is indeed statistical significance between the performances of detectors. When the detectors are rank- ordered by this act calculate, the order is generally comparable to those get hold of by other means, put it to some bodying that the arranging reveals authentic performance differences. SFOP was found to be better-quality to other detectors, while there are also some detectors whose performance differences were not statistically important. These findings are generally dependable with those get hold of by other investigators using unusual come within reach of, increasing our self-assurance that these concert differences are authentic. Researches were also try to get completed on stitching have common characteristics regions into landscapes, substantiating that enhanced reporting give ways a better quality consequence.