一种处理边缘不连续图像的雷达算法

Radar Algorithm of Processing Image with Discontinuous Edges

提出了一种处理边缘不连续图像的雷达算法.该算法的主要思想是,用统计学的方法判断一个像素属于一个区域的概率,从而把感兴趣的图像区域分割出来,即先记录像素点,包括点的颜色、坐标,再按一定的角度递增扫描图像边缘点以寻找边缘像素,其次把这些边缘点距扫描中心点的长度进行低通滤波以形成一个完整的边界,最后求出边缘线的质心,记录质心位置,并将这些数据作为自组织特征映射的输入且进行分类,由此完成区域的分割.该算法的时间复杂度为90×O(n3),较Hough变换的复杂度及鲁棒性更好.实验结果表明,所提算法在非规则区域分割上优于Hough变换,其运算效率高于视觉模型算法.

A new method named “radar algorithm” for processing images with discontinuous edges is proposed. The basic principle of the algorithm is as follows. It estimates the probability of whether a pixel belongs to a region of interest according to the statistics method to segment regions of interest in the image. Firstly, the coordinates and the color values of every pixel are recorded. Afterward, the current point is taken as the center point and the pixels are incrementally scanned in the boundary image with a certain angle to find a boundary pixel. Then all the boundary pixels are filtered through a low pass filter according to the distance between the center pixel and boundary pixels. The complete boundary is formed; the coordinate of mass center of the edge curve is computed and recorded. Finally, these values are regarded as the input of self-organizing feature maps and classified to complete the regions segmentation. The time complexity of the radar algorithm is 90×O（n^3）, and this new method has more advantages in spatial complexity and robustness over Hough transform. The experimental results show that the new method can create good results for regions segmentation. Furthermore, it surpasses Hough transform in irregular region segmentation, and more efficient than the vision model, especially in the processing of images with discontinuous edges. It can be extended for other applications.