Rapid Color-Based Segmentation in Digital Image Processing

A necessary part of digital image processing is segmentation of the images into a set of objects which exist on some background. We are interested in a class of objects whose distinguishing characteristic is their color. Such objects occur in many applications, such as microscopy, printing, production line monitoring, etc. In this work, a general method of rapid color-based segmentation is presented. The only hardware requirement is that look up tables (LUT) be available. Most image processing hardware contains either LUTs or processors capable of rapidly performing table lookup. The method presented allows simultaneous application of constraints on both hue and saturation. In addition, it allows for use of different color transformations. As such, it constitutes a general approach to analysis of images consisting of three spectral components. Because of the speed of LUT operations, this approach is suitable for many applications which are time sensitive. The major drawback of using LUTs is that the gray level resolution per color is limited by the size of the LUT. This method was implemented on a Matrox MVP-AT image processor, which is capable of processing 12 bit images (4 bits/color). In many cases, this resolution is adequate, as can be seen from examples which are presented.