Applications of Simple Markov Models to Computer Vision

In this report we advocate the use of computationally simple algorithms for computer vision, operating in parallel. The design of these algorithms is based on physical constraints present in the image and object spaces. In particular, we discuss the design, implementation, and performance of a Markov Random Field based algorithm for low level segmentation. In addition to having a simple and fast implementation, the algorithm is flexible enough to allow intensity information to be fused with motion and edge information from other sources