基于Shearlet变换的图像边缘检测

IMAGE EDGE DETECTION ALGORITHM BASED ON SHEARLET TRANSFORM

现存的基于空间域或小波变换域的图像边缘检测算法只能有效检测出有限方向的边缘,不能很好地检测较为复杂的边缘。Shearlet变换能够对图像进行任意层次和任意方向的分解,可以捕捉更多复杂的边缘。利用这些特点,提出一种新的基于Shearlet变换和改进canny算子的边缘检测算法。算法对canny算子进行改进,提出自适应阈值确定方法;根据Shearlet变换的多尺度和多方向特性,提出融合多个方向子图边缘信息的算法。实验证明,该算法不仅提高了边缘检测的完整性和精确性,还可以有效抑制噪声。

Existing image edge detection schemes based on special domain or wavelet transform domain can only capture the edges with limited directions but not good at detecting the complex edges. Shearlet transform can decompose the image at any level and in any direction, and can capture more complex edges. Using these characteristics,we propose a new edge detection algorithm which is based on the Shearlet transform and the improved canny operator. The algorithm improves the canny operator,moreover,the adaptive threshold determination method of is proposed as well. According to the multi-scale and multi-directional characteristics of the Shearlet transform,we propose an algorithm that fuses the edge information of multiple sub-images. Experiments show that the algorithm not only improves the integrity and accuracy of edge detection,but also suppresses the noise efficiently.