基于改进的Hough变换提取影像边缘

Image Edge Extractions Based on Improved Hough Transform

当影像中的边缘点较为密集或者噪声点较多时,直接应用Hough变换提取边缘线将会面临两个困难:一是虚假的峰值很多;二是边缘提取不准确。这两个问题往往会导致边缘提取的失败。针对这种情形,提出一种改进的Hough变换新方法,首先对分块影像类的边缘点进行Radon变换以获取局部直线特征;根据局部直线段的斜率、截距对其进行分类,同时计算出后续的Hough变换时所需要的ρ、θ取值范围,在此基础上进行经典的Hough变换。实验证明新方法具有很好的效果。

In remote sensing image processing, a lot of man-made objects can be described with the use of line features, such as build- ings, airports, roads, etc. Hough transform is relatively mature in many linear detection algorithms which have been proposed. However, when the edge points are relatively dense or much noise are contained in the image, direct applications of the Hough transform to extract edge lines will be faced with two difficulties： too many spurious peaks and high error rate, which will lead to the failure of the edge extraction. In view of this situation, an improved Hough transform method is put forward. Firstly, obtain local linear features of block image edge points with Radon transform. Secondly, classify them according to the slope and intercept of local straight lines, and at the same time calculate the ranges of ρ、θ, which are needed in the subsequent Hough transform. On the basis of this, do the classic Hough transform. Experiments show that the new method has a better result.