Performance Inspection of Gray-Level Difference Method Using Standard Images

Abstract

Pattern tracking algorithm using gray level difference as a similarity index of image fragments are inspected thronugh the application to the standard images which simulate images visualized with particles. It is shown that the performance of the algorithm decreases suddenly when the percentage of the particle area decreases to lower than 20%. The performance decreases also to the condition of high out-of-plane velocity, but the generation of incorrect vectors are suppressed in the successive abandonment prepared as a high-speed version of tracking algorithm, provided the percentage of particle area is high enough, due to the effect of statistical inspection involved in the algorithm. An idea to improve the quality of visualized images to prevent the decrease of tracking performance is also proposed.