基于双树复小波的图像边缘检测

Image edge detection based on dual tree complex wavelet transform

边缘检测在图像处理中具有非常重要的作用,它可以作为图像分割、模式识别及场景分析和图像检索的基础。经典的边缘检测算法相对比较简单,但方向适应性受到一定限制,尤其针对多方向的边缘,其检测结果明显不满足要求。在充分研究小波性能特点以及边缘检测的基础上,提出了方向选择性更强的双树复小波边缘检测算法。实验证明:该算法边缘定位效果好,边界相对清晰而且比较连续,同时能充分检测出图像细节部分的边缘。

Edge detection plays a very important role in image processing. It can serve as a basis for image segmentation, pattern recognition, image scene analysis and image retrieval. Traditional edge detection is comparably easy, but its direction adaptation is limited. Especially for multi-directories edge, the result of detection doesn＇t meet the requirement obviously. Based on a full research on the characteristics of discrete wavelet transform （DWT） and multi-directions edge detection, the paper proposed a novel edge detection scheme based on dual tree complex wavelet transform （DTCWT） which had stronger direction selection than DWT. The experiments prove that： By using the method, edge orientation is very exact and border is clear and continuous. At the same time, the proposed method can detect the edge of image detail.