Color image denoising using e-neighborhood Gaussian model

Takayuki Hara, Haike Guan

Abstract: This paper presents a novel denoising algorithm for color images. It is difficult to reduce color noise at high speed without losing image details. To solve this problem, the proposed method employs maximum a posteriori (MAP) estimation based on a Gaussian model in ε-neighborhood of the pixel and CIELAB color space. Using the correlation between RGB components in ε-neighborhood, color noise is reduced efficiently. Computational complexity is low because the method consists of non-iterative filtering and simple matrix operations. Experiments confirm that the proposed method preserves more image details, delivers PSNR close to state-of-the-art denoising algorithms, and involves less computation.