Abstract: Compared with wavelet transform, a drawback of Laplacian Pyramid（LP） in contourlet transform is implicit oversampling and non-orthogonal decomposition. This paper proposes an improved contourlet transform, the proposed method derives wavelet coefficients of entry image by wavelet transform, then reconstructs the image only by lowpass coefficients and after processing it can derive a highpass subband. The proposed method keeps critical sampling and it’s convenient for direction filter banks to partition the highpass subbands. It is found that the rate of matching in improved contourlet transform domain is higher than that in wavelet transform and contourlet transform domain.

Key words: image matching, contourlet transform, directional filter banks

摘要： 与小波变换相比，轮廓波变换中LP滤波器的显著缺点是存在过采样和非正交特性。提出一种改进轮廓波变换方法，首先对待处理图像实施小波变换，然后对低频子带图像进行重建，可得到一个高频子带图像，使用滤波器组进行多方向划分。该方法既保持了临界采样又利于后面使用方向滤波器组来对高频部分进行方向划分。实验证明该方法的匹配效果优于小波变换及轮廓波变换。

关键词: 图像匹配, 轮廓波变换, 方向滤波器组