基于非下采样Contourlet变换的多传感器图像边缘检测

摘要

A kind of multi-sensor image edge detection method based on nonsubsampled contourlet transform (NSCT) is introduced. The decomposition of the source images in a multi-scale and multi-direction way is realized by nonsubsampled pyramid (NSP) and nonsubsampled directional filter banks (NSDFB). The improved consistency checking strategy was used in the detection process: high frequency coefficients were processed using consistency checking not only in their own level but also in all high frequency levels. The nonsubsampled Contourlet transform has good multiresolution features of Contourlet transform and has translation invariance for image edge detection, and is applicable for handing multisensor image fusion. This method provides a new method for the image edge detection.%提出一种基于非下采样Contourlet变换(NSCT)的多传感器图像边缘检测方法,通过非下采样金字塔(NSP)和非下采样方向滤波器组(NSDFB)实现了对图像的多尺度多方向分解.在检测处理中.采用改进的一致性校验策略:高频系数除了进行本层的一致性校验外,还进行多层之间的一致性校验.非下采样Contourlet变换保留了Contourlet变换良好的多分辨率特性,又具有平移不变性,更适应于进行多传感图像融合,为图像边缘检测提供了新的方法.