Development of a new detection method of image primitives by several procedures in image space and parameter space

点击翻译按钮获取中文摘要

In image analysis, it is important to detect straight lines, circles and ellipses which exist in the image and are called image primitives, because these image primitives represent edges of the objects in the image. The most popular technique to detect the image primitives is Hough transform. At the transform, resulting peaks in the accumulator array which are gotten by a voting procedure in the parameter space represent strong evidence that a corresponding the image primitives exist in the image. In the voting procedure, a large number of votes are necessary and it makes the transform slow. In this research, we developed speed-up methods which reduce the parameter space and remove the unnecessary vote by introducing several procedures at the image space. From the experimental results and theoretical analysis, we confirmed that the developed methods have good performance.