The Research On High Density Salt And Pepper Noise Image Restoration Methods

Image is the objective visual reflection of nature things, and it is an important carrier for information acquisition and dissemination.With the rapid development of computer and digital technology, digital image has become an important way for visual information storage and transformation. However, in the process of digital image acquisition, transmission and storage, it is inevitably affected by noise pollution, which causes the image quality degradation.In order to restore the true image, noise image must be repaired. Then, image denoising is a hot research topic in the field of image processing.The salt and pepper noise look like white or black particles can severely degrade image quality. Especially, the high density impulse noise will make noise image can not been recognized. Recently, the switch median filter is adopted for salt and pepper noise image restoration and achieved good effect in low noise density. However, with the increase of noise density, the repair effect of this algorithm is poor. This paper aim at the high density salt and pepper noise image restoration problems, carry out the following research works:1) Neighbor adaptive switching median filter:To solve the image blurring problem of switch median filter caused by using a fixed and large neighborhood size, a neighbor adaptive switching median filter is proposed. It take the non pollution pixel number as a criterion to chose the neighborhood size step by step increasing, and replace the polluted pixel with the median of the non-polluted pixels. Experiments show that the algorithm can significantly solve the image blurring problem, and completely eliminate salt and pepper noise under arbitrary noise density (1%-99%). The image inpainting quality is obviously superior to the switching median filtering method.2) Switch local extreme median filter:To solve the problem that the median estimation is not close to the polluted pixel truth value caused by the extremely less non polluted pixels in high noise density images, the switch local extreme median filter is proposed. It use the maxima and minima of non polluted pixels in adjacent neighborhood to clamp the polluted pixels estimation ranges, and take the mean of these extreme and non polluted pixel in current neighborhood to replace the polluted pixel. The image restoration effect, especially in image details, is better than switching median filter and the neighbor adaptive switching median filter algorithm.3) Switch least square filtering:To solve the problem that the median filter is not use the image smooth characteristics, we consider an image as a 2D linear system. The spatial pixel position as input, and the pixel value as output. In switching median filter framework, the populate pixels are restored with this linear model according on its pixel location with a least squares estimator. Experiments show that the proposed method can significantly improve the quality of image restoration in high density noise pollution.In this paper, three kinds high density salt and pepper noise image restoration methods are proposed in switching median filter framework. The experimental results show that they can meet the high density of salt and pepper noise image restoration application requirements and the performances are better than existing salt and pepper denoise methods.