Research On Image Enhancement Based On JND Curve Property

During acquirement, transport and digitization of images, the descend in the quality ofimages is inevitable. Image processing with the algorithm of image enhancement can improvethe quality of image much, and increase the details of image and visual effect. Enhancementtechniques is important for image preprocessing and make the processed image have the extraadvantage which compared with the original image in the region of the parameter estimation,image segmentation, etc. The human visual system are on the basis of processed image.Traditional technologies of image enhancement are based on digitization, they don’tconsider visual effect directly. Processed image increase visual effect but amplifier noise atthe same time, the defect comes from the global gray level processing. The paper based onhuman visual characteristic presents an algorithm of image enhancement which calculatesdifferent pixel in different parts, and improves the quality of image for processed image havehigher visual effect, no extra noise and more information of image. The relative detailed workcan be concluded as follows:First of all, conventional average filter can not distinguish the point of information andthe point of noises, and use the same way for all the point of pixels, to make the image blurryafter de-noising. Focus on the defect, the paper presents an algorithm to reconstruct imageswhich uses just noticeable difference to distinguish pixel points and noise points, differentkinds of points given different weights.Secondly, the algorithm adjusts the image pixel to the whole gray level. Refer to justnoticeable difference curve and Weber’s law, it uses the peak and the maximum gray of lowarea in the histogram, to find suitable threshold, compressing high gray area, expanding lowgray area, to enhance the average brightness of image, and to make observer can analyze theprocessed image in a better light source.Finally, focus on the defect which equal grays of image changed at the same time in thetraditional the algorithm of image enhancement, the paper presents a better algorithm whichcalculates different pixel in different parts. In matrix, based on advanced just noticeabledifference curve, the paper makes indistinguishable gray incorporation for the better visualeffect between the combined gray and next one. After the emulational experiment, it showsthat the present algorithm enhances contrast of image relatively, improves the sharpness ofdetails and increases information, for the processed image fits with the human visual system.