Edge Sharpening of Digital Images using Edge Profile Analysis

Abstract

In obtaining a digital image, image acquisition imperfections produce edges that are blurred. Consequently, edges are more closely modelled as having a ramp profile. For scientific purposes, sharp images are required. Conventional methods of image enhancement have mainly attempted to increase contrast level of the image, which results in a better appearance of elements, regions and different objects in an image. Edge enhancement is one of the ways of sharpening a blurred image. Edge detection is the most common approach for detecting discontinuities in grey level of an image. If the transition from one region to the other over an edge is clear and the change is abrupt then the difference between the two regions would become clearer and thus an enhanced image.

The slope of the ramp is inversely proportional to the degree of blurring at the edge. If the steepness of the slope can be increased from a ramp profile to a step edge profile then there would be a reduction of the blurring. Using the local analysis of gradients, ramp pixels can be detected. Sobel operator is used for this purpose. Throughout the ramp region, the first derivative of the image function is positive. At the boundary of the edge, the first derivative is zero. If the ramp region is detected from the middle of the ramp, where the first derivative is maximum, then the ramp pixels can be interpolated with the intensity values of the relevant border region pixels.

The technique is used to analyse images with different properties like edge angle, intensity difference of the two regions, edge strength and edge width. Results are also compared with the results obtained with Adobe Photoshop. Noisy images and images with text are also enhanced. A review of a research already done on edge enhancement critically analyses the technique.

The technique is simple, robust and based on the fundamental properties of an edge. It is computationally inexpensive and achieves the objective.