A Neighborhood Decision Based Mean Filter for Video Image Denoising
Qian Xu, Xiang Ji, Che Qu · 2023
Denoising Gaussian noise generated in digital images during acquisition and transmission requires balancing the denoising effect with the loss of picture details, and existing spatial domain filtering methods have their advantages and disadvantages. In this work, we propose an improved method based on mean filtering, which uses a depth-first search algorithm to find connected blocks with similar colors to the center point of the current filter window, and only counts the values of these pixel points for the filtering operation. This method improves the inherent defects of mean filtering by considering the correlation of colors between pixel points. Moreover, we conducted an experimental design considering the acceptance characteristics of film and television viewers. The quantitative and qualitative results of the experiments show that our method has certain advantages in preserving the edge clarity of color blocks and the sharpness of the image when applied to the noise reduction of film and television image frames.