A system for sequential step detection with application to video image processing

A method for detecting the occurrence of an abrupt steplike change in a time sequence of video images is presented. A single-pole recursive high-pass filter cascaded with a moving average filter processes the input data to remove the quiescent background level and accumulate a sustained change in amplitude. The absolute value of the output is compared to a threshold to decide whether a steplike change in signal amplitude has occurred. It is shown that, for a given cutoff frequency of the high-pass filter, an optimal value exists for the number of terms in the moving average. Considerations for implementation of the algorithm on practical image processors are discussed. The results of numerical and laboratory experiments are presented that verify the effectiveness of the method