分割双纹理图像的最佳Gabor滤波器设计方法

摘要

A new method for designing optimal Gabor filters is proposed, which aims at segmenting bi-textured image with a single Gabor filter. It is known that the energy distribution of texture is centralized relatively in the frequency domain, but not focused on a single frequency component. Therefore, the optimal Gabor filter is designed by comparing the Fourier power spectrum density between two types of textures. Compared with other existing algorithms,the proposed algorithm requires less computational load and acquires better segmentation result. Moreover, a strategy to search the optimal filter with the help of the bandwidth parameter of Gabor filters is provided, which is matched with the perceptual property of human vision system and has great maneuverability.%本文提出了一种用单Gabor滤波器分割双纹理图像时最佳Gabor滤波器的设计方法。该方法根据纹理图像能量在频域上的分布相对集中但并不局限于单一频率的特性，通过比较两类纹理之间的Fourier功率谱密度来设计最佳Gabor滤波器。与现有的方法相比，它具有计算量小、分割效果好的特点。另外它还提供了一种根据Gabor滤波器的频率带宽参数来指导搜索最佳滤波器的新策略，这与人类视觉系统的特性是相吻合的，具有很强的可操作性。