论文部分内容阅读
Edge-preserving multiscale representation of images has many applications in computational photography and other fields of image or video processing.We present a novel image smoothing scheme for multiscale image decomposition, which is based on the fact that several simple and popular filters can smooth images efficiently except near edges where images can be blurred.The image smoothing scheme consists of two steps: (1) an image is smoothed by some known smoothing methods with no guarantee of edge preservation;(2) the difference image between the smoothed and the original images is compressed nonlinearly and the blurred edges are recovered.This smoothing method owns the properties of simplicity, easy implementation, linear time complexity and low memory spending.It is very suitable for recursive image smoothing or decomposition.The examples we have experimented show that our proposed algorithm can achieve similar or even higher quality results as compared with other known image decomposition algorithms, but the speed has been improved tens to hundreds times faster.Due to its efficiency and high speed, this new image decomposition method is also applicable for interactive photo or video editing by which the effects are controlled by a few parameters selected by users.