Object Boundaries in Digital Images

Abstract

Determining the boundaries of objects from digital images, which is important for their measurement, is often a challenge. Limitations are imposed by nonuniform lighting or density, optics, dimensions of solid-state detector arrays, noise, and pixelation of the image. A wide variety of computer algorithms illustrated here seek to address these problems, but there is no universally optimum solution. Often, the goal is to achieve repeatability on a particular class of images and objects rather than absolute accuracy.