|
The grey system theory has recently been emerged as a powerful tool for image processing, image analysis and image
understanding, such as, image compression, image denoising, image edge detection, image hiding, objects recognition
and classification, image retrieval, image fusion and so on. However, these are mainly analyzed based on the
technologies of grey model and grey relational analysis. The grey difference information principle is one of six basic
principles of grey system theory. All there must be the difference in information, namely, the difference is the
information. But so far the studies in image engineering based on this principle are rare. And image noise is exactly
shown as the difference of image information. Therefore, the images are always corrupted by noises, which have a bad
influence on the subsequent processing. Though some classical filters have successfully been used in gray scale imaging
to remove impulsive noise, their extension to color images is not direct. The main difficulty is that an order has to be
defined to sort the color vectors. In this paper, the grey difference information principle is applied to a novel valuable
color image denoising strategy. Firstly, the grey difference information principle is introduced in detail including
difference information sequence and difference information measure. Secondly, the basic idea of color image denoising
is proposed based on the difference information principle, while the pixel difference information sequences are
established in the light of the template types of filters and the properties of image noises. The novel method will select a
suitable pixel, which has a minimum difference information synthesis measure in the predefined template window, to
replace pixel of the noise. Lastly, the experiments are analyzed based on the novel method and the classical filters. The
experimental results demonstrate that the proposed method outperforms the conventional mean filter in removing
impulsive noise in color images. The successful application indicates that it is just as feasible and effective to use grey
difference information principle as the technologies of grey model and grey relational analysis to process the color image.
|
