Moving Object Area Identification in Image Sequence

The paper carries out a comparative analysis of various approaches to identify area of moving object in image sequence using pixel-by-pixel estimation of inter-frame shifts. It studies projections and polar parameters of shift vectors of reference image points corresponding to the nodes as their estimated parameters. The paper proposes two methods for estimating shift vectors' field. The first method uses stochastic gradient descent algorithm to sequentially process all nodes of the image row-by-row. It processes each row bidirectionally (from the left to the right and from the right to the left). The joint processing of the results allows compensating inertia of the recursive estimation. The second method uses the correlation between rows to increase processing efficiency. It processes rows one after the other with change in direction after each row and uses obtained values to form resulting estimate for each node. The paper studies two criteria of its formation: gradient estimation minimum and correlation coefficient maximum. It also studies computational cost of the proposed algorithms.