Research On Filtering Algorithm For Salt-and-pepper Noise Of Digital Image

Digital image is often interfered by many kinds of noise in the process of image production and transmission, which has strong impact on the quality of image. Therefore it is necessary to filter the noise before edge detection, image segmentation, feature extraction, pattern recognition and other digital image manipulation work. And one of the hot spots in this area is how to guarantee the most perfect image details through the filter process.Salt-and-pepper noise, exists in digital image extensively, can severely damage the digital image. And the Median filtering, which we usually use as traditional method, is a non-linear filtering methods. It can restrain salt-and-pepper noise at a certain extent but it is also blindfold to execute the same way to process all pixels. It would change the pixel of some signal spot during the filtering and cause a large loss of the image details.Based on the systematic research of noises species and characteristic, this thesis expounds the major study on salt-and pepper noise filtering algorithm. Please find the ground work and achievement as below:1.It studies on the basic characteristic of digital image and filtering algorithm.2.Median filtering algorithm uses a unified way to process each pixel. Based on this weakness of median filtering, it proposes a new filtering algorithm with accurate noises detection step. The algorithm separates the noise and signal through setting threshold value and considering the relativity of adjacent pixels. It can keep the original gray value and filter out the noise to protect the image details and increase the filtering precision.3.It makes a research on GF(28 ) field and some related questions. According to the relativity of adjacent pixels, it structures a similarity function in GF(28 ) field and makes a further point of image adaptive salt-and-pepper noise detection operator and an improved filtering algorithm with accurate detection capability for noise. This algorithm is accurate on noise detection and advanced in image restoration and image details protection.