Concept of Pseudo-afterimage and its Application to Moving Objects Detection

Abstract

A concept of pseudo afterimage which expresses the movement information of moving objects in a stationary image is proposed to treat(analog) afterimage phenomenon (in the human perception)by digital dynamic image processing. An efficient construction algorithm of the pseudo afterimage is presented and it is applied to the moving objects detection. The method accepts the pseudo afterimages from a series of digital dynamic color image as an input, and uses the fuzzy inference based on 15 if-then fuzzy rules that express the relations between each color ingredient and the movement of moving objects. Furthermore, a gain filter as the pre-processing of fuzzy inference is proposed to suppress a background moving effect. The problem that a position gap occurs in conventional methods using difference images or correlation calculations is solved by the proposed method. Detecting the vehicle running on general road from dynamic images taken from a camera on a car is performed as the basis of future ITS recognition experiments.