摘要： 在轮廓波变换中进行拉普拉斯金字塔分解时，所得的带通图像在奇异性点附近产生振荡，影响了图像去噪的效果。针对该问题，提出一种改进的拉普拉斯金字塔分解，可消除边缘附近的震荡。利用改进的拉普拉斯金字塔实现轮廓波变换，并对图像进行自适应去噪。实验结果表明，该算法所得的峰值信噪比较轮廓波变换自适应去噪算法有显著提高，且视觉效果有较大改善。

关键词: 轮廓波变换, 拉普拉斯金字塔, 图像去噪, 奇异性

Abstract: In the contourlet transform, the image obtained by Laplacian Pyramid(LP) decomposition may produce artifacts on singularity of signal, which is harmful to image denoising. Due to the lack, a new LP decomposition is proposed and it can suppress the artifacts around the edge effectively. The new LP is used to implement the improved contourlet transform. Denoising experiments for Lena image using adaptive threshold show that denoising performance of the improved contourlet transform outperforms the contourlet transforms both in peak signal-to-noise ratio and visual quality.

Key words: contourlet transform, Laplacian Pyramid(LP), image denoising, singularity

中图分类号: 

• TP391