采用奇异值分解的图像模糊区域检测分割新方法

An Image Blurred Region Detection and Segmentation New Method Using Singular Value Decomposition

针对局部模糊图像的模糊区域检测分割问题,提出了一种改进的基于奇异值分解和图像抠图的模糊区域自动检测分割算法。首先,采用分块的方法对局部模糊图像进行再次模糊,通过比较前后图像块的奇异值特征变化差异将其标识模糊块或清晰块以得到一个标识图。其次,根据标识图,结合图像抠图技术对图像的局部模糊区域进行自动提取。实验结果表明,该方法可以较为精确地检测并分割出局部模糊图像中的模糊区域。

Partially blurred images caused by motion or out of focus are common in life, we need to detect and segment blurred image region when doing some multimedia analyzing tasks. As for the problem of blurred region detection and seg- mentation of partially blurred images, we propose a method based on singular value decomposition and image matting is presented in this paper. Making use of the different changes of image blurred regions and clear regions under a low- pass fihering, partially blurred image is firstly partitioned into patches and each image patch is re-blurred by a Gaussi- an function, then a trimap is got after comparing the singular value feature difference of the patches to distinguish blurred regions and clear regions. Secondly, a image matting technique is combined with the trimap to automatically segment blurred regions from the partially blurred images. Experiments show that this method can detect and segment the blurred regions accurately.