Research on introducing a factor K self-adaptive smoothing algorithm in image processing

For an image accompanied by Gussian noise, image smoothing is usually achieved through adopting the classical Gaussian smoothing mask. This may sometimes blur the edges and other fine details. Noise removing and edges blurring are two conflicting requirements. And delta, as a key factor in the Gaussian function, can greatly affect noise removing, edges blurring and even brightness of an image after processing. In this paper, the theory of Gaussian smoothing in continuous domain is analyzed. With an adjustive factor K being introduced, an improvement on the Gaussian Smoothing algorithm is described. We try to find out the relation among the value of delta, the value of K and the result of image smoothing. Experiments confirm the proposed improvement. The relation gives us an indication that we should select appropriate values of delta and appropriate values of K according to the values of SNR in different regions of an image, then establish smoothing masks to obtain the best result of image smoothing. Some ideas about the new self-adaptive smoothing algorithm are presented simply. By employing the self-adaptive algorithm, the results of processing an image accompanied by Gaussian noise are shown. Finally we should conclude that we can obtain fine details remaining in some partial regions at the cost of a little reducing of the SNR character in the whole image after the algorithm being used.