Edge preserving image enhancement using median absolute deviation and iterative median absolute deviation

Abstract

The aim of image enhancement is to produce a processed image which is more suitable than the original image for specific application. Application can be edge detection, boundary detection, image fusion, segmentation etc. In this paper different types of image enhancement algorithms in spatial domain are presented for gray scale images. Quantitative analysis like AMBE, MSE and PSNR for the different enhancement algorithms are evaluated. Algorithms like Linear contrast stretching, Fuzzy based image enhancement, and Local mean local standard deviation are discussed and compared. During result analysis, it has been observed that some algorithms do give considerably highly distinct values of PSNR, MSE and AMBE (parameters) for different images. To stabilize these parameters, a new enhancement scheme Median Absolute Deviation (MAD) is proposed, which will take care of these issues. By experimental analysis It has been observed that proposed method gives better AMBE (should be less) and PSNR (should be high) values compared with other algorithms, also these values are not highly distinct for different images. Also elapsed time taken by different algorithms to achieve desired application is recorded. MAD based image enhancement is faster also provides best values of MSE, PSNR and AMBE when applied on a database consisting of 400 images.