Research On Novel Filtering Algorithm For Noise Deduction From Color Image Based On Peer Group Concept

During the process of acquisition and transmission, image is easy to be susceptible to noise interference and pollution, not only seriously affect ing the quality of the image itself, also hindering the observation of the image and recognition. Along with the rapid development of computer and multimedia technology in recent years, especially the constant progress of image imaging equipment for high quality color image provid ing a good hardware support, the applications based on color image are becoming more and more popular, and the related image processing technology has drawn more and more attention.The basic task of the color image denoising images is to effectively filter out the noise pollution of attacks on the images, while protecting the overall image quality, including the edges and detailed areas. According to the origin, the image noise can be roughly divided into two types, namely, impulse noise and gaussian noise. Face to the characteristics of image noise, kinds of image filtering algorithms have been widely proposed and applied, not only improving the quality of the image well, but also provid ing a good theoretical basis for the research and design of the new filter algorithms.This paper first launches the research on color images polluted by impulse noise. Through the traditional study of vector median filtering algorithm, refer ring to the latest modified double step peer group filter algorithm is put forward, the performance of the algorithm is improved with the concept of peer group, including the preview work before noise detection phase and applying the concept of aggregated distance during the noise filtering phase, which has well solved the problem about preserving the edges and details of color images under relatively high impulsive noise density.Hereafter, this paper extends the circumstance of image contamination to mixed noise level, and designs a novel image filtering algorithm, employing the concept of fuzzy peer group for more complex image processing conditions. Specific measures have been taken by further appling the theory of fuzzy sets to fully extend and optimize the noise detection and filtering part based on the basic structure of the previously proposed algorithm above. The novel algorithm can accurately determine what type of noise has been exactly added to certain image pixels, including impulse noise, gaussian noise and even mixed noise interference, and then adopts corresponding filtering method according to different circumstances.Finally, in order to make the algorithm better meet human visual sensory experience, this article extends the algorithm from the traditional RGB color space to the CIELAB color space, which is more adapted to human visual observation. From the results of the simulation and the restored image, it can be clearly observed that the novel proposed algorithm has obtained better filtering effects than others.