Circles Detection in Images by Using of Coarse-to-Fine Search Technique

In the automatic production, the application of machine vision technology on image pattern recognition is an important task. Realistic image is composed of these basic units such as straight lines, circles and ellipses. But the actual images usually contain noise and other interferences. So the images will have some discontinuous modes. Hough transform is commonly used to detect straight lines, circles or other parametric patterns in noisy images. In practical application, Hough transform requires a large amount of storage space and computation. In the paper, it proposed an efficient coarse-to-fine search technique to reduce the storage and computing time of circle detection in image. Variable size image and accumulated matrix are used to reduce the required amount of computation and storage for Hough transform. It shows the parameter convergence speed and precision by using different iterative algorithms. The experimental results show that the coarse-to-fine search technique is very suitable for circle detection having time constraints.