基于模糊相似度融合的图像复原算法

Image Restoration Algorithm Based on Fusion with Fuzzy Similarity

传统的图像复原技术主要是通过单一的方法对图像进行复原,而图像融合技术主要集中于对多聚焦图像融合和多源图像融合.文中从图像融合的基本思路出发,基于局部特征信息的模糊相似度的融合策略将多幅复原图像进行融合,提出基于模糊相似度融合的图像复原算法.首先依据区域特征(方差、梯度和均值)确定模糊相似度,然后将具有一定相似性的复原图像进行基于像素点加权平均,权系数由该像素点处的区域特征确定,最后获得融合图像;对不具有一定相似性的复原图像按照"最大值选取"原则,依据"区域特征较明显者提取像素点"来获得融合图像.实验结果表明,该算法复杂度小,融合后复原图像改善的信噪比比融合前提高了0.1～1dB;同时,比经典算法的效果有明显改善.

Traditionally image restoration technology is mainly focused on restoring the image with single method. Image fusion technology is applied on multi-focused images fusion and multi-source images fusion. Inspired by fusion strategy, some restored images are fused into one image based on fuzzy similarity. So an image restoration algorithm based on fusion with fuzzy similarity is proposed. First, compute the fuzzy similarity based on region character （gradient, variation and mean）, then some restored images which have certain similarity are weighted based on pixel, and some another images having not similarity are obey ＂maximum value selection＂ principle and the final fusion image are determined. Experimental results show that the improved signal noise ratio （ISNR） of fusion image is 0.1-1 dB better than before. ＂At the same time, the results are better than traditional method.