A novel technique for detecting edges in a homogeneous background

This paper proposes a novel technique for detecting the edges of 2 1/2 -D planar and multi-surface objects placed in a homogeneous background, i.e. a background that has the same grayscale intensity as the object surface. In such applications, existing intensity-based methods of edge detection are unable to detect the edges successfully because a variation in intensity does not exist at the edge pixels. In the proposed technique, sinusoidal fringes were projected obliquely onto the object surfaces to form a fringe pattern image. Due to the depth variation at the various surfaces of the object, intensity break points occur at the edges. The edges were then detected by locating these intensity break points in the image. Analyses carried out on raw fringe pattern images in simulation studies, using the proposed technique, show that the accuracy of the edge detection decreases with increase in the standard deviation of Gaussian noise. However, analysis carried out on a phase image, obtained by applying the phase-shift method, shows a significant improvement in accuracy. The results of edge detection on real images are presented.