An Empirical Study Using Line Profile Histogram Approximation of Edge Detection Algorithms

Abstract

Traditional edge detection algorithms such as Sobel and Prewitt have been used in many edge detection applications. Canny have proved to be quite efficient in detecting thin edges. This paper discusses the efficiency of these three techniques in the detection of the cortical outline of long tubular bone from radiographs images. The method involves the use of line profile histogram approximation (LPHA) algorithm to detect the outline positions. These positions are then used to calculate geometric measurements such as the internal diameter (ID), outer diameter (OD) and the cortical thickness (CT). The accuracy of these measurements are compared with its retrospective manual measurements which was measure using micro calipers by Dr. Lee Cheng Wai. Visually all three techniques performs reasonably well in detecting the bone cortical edge. However in the quantitative measurements, Canny perform quite well in certain cases but not in others. The performance of Sobel and Prewitt are consistent in most cases but less accurate as compared to Canny.