关键词: 图像复原, 图像去噪, 空间结构, 自适应, 专家场

Abstract: Image restoration is an important research subject in image processing.Image gray or gradient magnitude is used in most of image restoration algorithm and spatial information is not adopted.A Field of Experts（FoEs） model which Roth et al presented adopts some fixed filters by learning.These filters can throughly characterize image spatial information，but it is not adaptive.In this paper，a spatially adaptive FoEs model is presented，which can adaptively adjust own filters to image spatial information.In FoEs model，filters are designed according to image edge direction and can be adaptively adjusted.When the new FoEs model is used for image denoising，varying restoration processing is adopted according to image spatial information.The results show that the new FoEs model significantly improves the image restoration quality and its performance is over traditional FoEs model’s.

Key words: image restoration, image denoising, spatially structure, adaptive, field of expert