基于约束抽样Hough变换的圆轮廓快速检测技术

针对Hough变换进行圆轮廓测量存在的问题，提出了一种新的Hough变换圆轮廓检测方法．该方法综合利用图像本身灰度信息和梯度信息，对图像空间的抽样进行约束与优化，减小参数的累加数目，得出图像空间中属于目标的轮廓点；并根据目标本身的性质进一步搜索，去除伪轮廓点；采用了“多对一”映射与并行算法来提高检测速度．实验结果证明应用本方法对于对比度低、有噪声污染的图像同样可以检测出真实轮廓，定位不确定度〈0．15像素，运算速度是传统随机Hough变换的2倍．

Fast circle contour detection technique based on hough transformation with sampling restraint

Hough Transform(HT) is widely used in computer vision by its transformation from image space detection to parameter space accumulation.A new algorithm for detecting circle contour based on HT is proposed.Using this method,the information of grayscale and gradient is used simultaneously, the sample in image space is constrained and optimized,the amount of accumulator is minimized,the contour point belonging to the object is attained,the dummy point is picked out by re-searching contour according to the geometric property of object,the computing speed is accelerated through multi-one mapping and parallel algorithm.The result of experiment shows that the algorithm can be used in the image with low contrast and spoiled by noise,the uncertainty is less than 0.15 pixel,and the computing speed is twice comparing the classical Random Hough Transform(RHT).