A BPEJTC-based segmentation for a non-stationary image

This paper presents a correlator-based segmentation algorithm applied to a non-stationary image. We show that in this algorithm, a moving target's signature residing in the image sequence can be extracted via monitoring the levels of correlation peaks between two adjacent images. We also present an electro-optical system to implement the algorithm. It is shown that the performance of this method depends to a great extent on the correlation results and thus the proposed system uses a non-linear JTC called BPEJTC that has a much higher peak-to-sidelobe ratio than a conventional JTC. Moreover, we show that this system effectively distributes the required computations over an optical unit and a digital unit to reduce the processing time, and the inclusion of optical processing enables the system to work well even for massive input data size. Experimental results obtained with a real image sequence are demonstrated.