用于图象分割的冗余小波包方法的研究

A Study of Segmentation with Redundant Wavelet

采用微特征对纹理图象进行分割时涉及局部窗口的选择问题，如果窗口太大会引起边界不清，太小又使得统计性能不稳定，尤其采用传统的小波包分解，随着层数的增加窗口会变得越来越小，为此引入冗余小波分解，分解时各个子图都与原图大小相同，实验结果表明采用冗余小波分解明显改进了分割性能．本文也对能量特征和熵特征做了比较研究．

Local window selection is involved in texture seg mentation with microfeatures . Boundblur arises out of bigger window ,smaller window causes statistics performance unstable . When tra dition wavelet is used ,the window beco mes more and more smaller with levelincreasing . So redun dant waveletis introduced . Subimage is as sa me size as original image . Experimental result showsthat it im prove seg mentation performance . Energy feature and entropy feature com position is madein this paper .