Using Vector Quantization of Hough Transform for Circle Detection

Circles are important patterns in many automatic image inspection applications. The Hough Transform (HT) is a popular method for extracting shapes from original images. It was first introduced for the recognition of straight lines, and later extended to circles. The drawbacks of standard Hough Transform for circle detection are the large computational and storage requirements. In this paper, we propose a modified HT called Vector Quantization of Hough Transform (VQHT) to detect circles more efficiently. The basic idea is to first decompose the edge image into many sub-images by using Vector Quantization algorithm based on their natural spatial relationship. The edge points resided in each sub-image are considered as one circle candidate group. Then the VQHT algorithm is applied for fast circle detection. Experimental results show that the proposed algorithm can quickly and accurately detect multiple circles from the noisy background.