Overlapping Circles Detection

As part of work into an effective means of detecting blood cells in microscopic images, a novel circle detection technique was developed. Utilising desirable properties of the Gabor wavelet, a filter called the Gabor Annulus technique was proposed, offsetting the traditional Gabor filter by an origin-wrapped radius. The method was designed to accomodate the existence of extensive noise and overlapping circles, and was tested on synthetic and real images, while being compared with existing methods including the Circular Hough Transform. Results show a significant improvement on previous approaches, with higher mean response ratios in all cases and clearer responses in high level noise. In the case of real test images, the Gabor Annulus technique demonstrated the best detection rates, with the lowest false positive rates.

Citations: [·]
Last updated 2014 | School of Computer Science