波原子纹理图像阈值算法.pdfVIP

第 31 卷第 8 期 电 子 与 信 息 学 报 Vol.31No.8 2009 年 8 月 Journal of Electronics Information Technology Aug..2009 波原子纹理图像阈值算法 刘国军①② 冯象初① 张选德① ①(西安电子科技大学理学院 西安 710071) ②(宁夏大学数学计算机学院 银川 750021) 摘 要：该文提出了一个新的依赖于 Besov 光滑参数和尺度的波原子软阈值纹理图像去噪模型。该模型充分考虑 了新的多尺度几何分析工具——波原子的优良特性，比如波原子正交基和框架的灵活选取，对振荡纹理图像的稀疏 表示，波长和支撑尺寸满足抛物尺度关系等。数值实验表明，新模型比硬阈值和软阈值具有更好的去噪性能，而且 随着 Besov 光滑参数的增加信噪比有明显提高。 关键词：图像处理；波原子；纹理；Besov 空间；光滑参数；阈值；尺度 中图分类号：TN911.73 文献标识码：A 文章编号：1009-5896(2009)08-1791-05 Threshold Algorithm of Texture Images with Wave Atoms ①② ① ① Liu Guo-jun Feng Xiang-chu Zhang Xuan-de ① , (School of Science, Xidian University, Xi an 710071, China) ② (School of Mathematics and Computer Science, Ningxia University, Yinchuan 750021, China) Abstract: A novel denoising model for texture images is proposed, which is the soft threshold algorithm depending on both the smoothing parameter in Besov spaces and the scales of wave atoms. This model well considers the good properties of new multiscale geometric analysis tool—wave atoms, such as the flexible choice of the orthonormal basis and tight frames, sparse representation of the oscillatory texture images, as well as parabolic scaling between wavelength and the size of the essential support. Numerical experiments show that the proposed model not only has a better denoising performance comparing to the hard and soft threshold, but also significantly improves the SNR with the increase of the smoothing parameter in Besov spaces.