Research On Compressed Image Mosquito Noise Reduction Method

With the rapid development of science, technology and the Internet, people's daily lives and digital images are closely related. Because digital images have enrich data, images have to be compressed encoding before they are transfered. In order to achieve image compression and image transmission in limited bandwidth, scale transformation and image compression should be applied in digital images in the DCT domain. Now, there are many image or video compression coding standards, such as MPEG, JPEG and H.26X series and so on, and the most of them adopt the form of 'block DCT+ quantitative'. They can execute compression coding for images effectively. Although the image compression reduces the size of the images, it also leads to a series of distortion phenomena, such as ringing noise, mosquito noise, block effect, stair stepping effect and so on.Today, people have higher requirements for images, and they are eager to have clearer images which lay the foundation of handling images further. Image denoising has always been the focus of the digital images processing field. Due to different characters from all kinds of noises, we can't only use one uniform noise reduction algorithm. Compared with other noises, there are limited literatures about the suppression of mosquito noises. For this kind of special reduction processing of mosquito noises in compressed image, there are some studies in this thesis.With the development of digital image processing techniques, the methods and theories of image denoising are also enriched and developed constantly. However, some classical denoising filtering algorithms such as mean filtering,median filtering, and gaussian filtering have some inhibitory effects on noises, and these algorithms will cause the loss of images' details and edges informations when they are used to filter noises. Aiming at the coding for image mosquito noise reduction, we introduce the idea of edge detection and image local filtering. Based on the spatial characteristic of mosquito noises, we improve traditional ?-filtering algorithm which is used to remove mosquito noises, and propose adaptive ?-filter which based on a local variance. Compared with traditional s-filtering algorithm, adaptive ?-filtering algorithm is better in removing the mosquito noises of different intensity and maintaining the edges of images. In order to achieve the purpose of reducing mosquito noise better and preserve image details, we apply the bilateral filtering algorithm on the compressed image of mosquito noise reduction and change traditional bilateral filter's entire blind filter into the filter strength is adjusted adaptively according to the local image. By improving the traditional bilateral filtering, a method based on gray level difference adaptive bilateral filtering for removing mosquito noise is proposed. The results of experiments show that this way not only makes up the fuzzy defects of edges and details in traditional blind filtering for the full images, but also removes noises better and retains details information.It is successful to avoid the emergence of fuzzy phenomenon after processing images.