A Quantitative Performance Analysis of Edge Detectors with Hybrid Edge Detector

Abstract

Understanding of images via features like edges plays a vital role in many image processing applications. However, obtaining an optimum edge detector that performs well in every possible imaging condition is still an open challenge to researchers. In this paper, a quantitative analysis of some significant state of art edge detection techniques such as Canny's, Prewitt, Sobel, Laplacian of Gaussian, fuzzy based edge detection, wavelet based edge detector with hybrid edge detection technique is proposed based on the correspondence between their outcomes. The hybrid edge detection method utilizes fuzzy logic partitioning along with wavelet transformation to maintain a proper balance in the false detections i.e., false positives and false negatives rates and provides better tracking of edge information. Various subjective as well as objective quality measures are provided for quantitative analysis of edge detectors. The experimental results confirm that compared to other techniques the hybrid edge detection technique outperform in terms of edge detection accuracy exclusively when the images are corrupted by noises.