Displacement Estimation for Objects on Moving Background
In a sequence of TV-pictures moving objects can be detected by their frame-to-frame displacements. In the case of a moving TV sensor also the non moving background will show a frame-to-frame displacement. Here a method is presented which allows the estimation of the displacement relative to the background and an automatic detection of relatively moving objects. The main problem is the estimation of the background displacement. A method is described to estimate the background displacement with low computational complexity for a background displacement varying spatially in the picture. For simplification the assumption of a flat surface of the ground and a small size of relatively moving objects has been made. Two laboratory generated sequences have been used to determine the accuracy of the background displacement estimate and the amount of relative displacement needed to separate moving objects from background. The error of the estimated background displacement shows a standard deviation of 0.50 pel per frame if only background motion occurs. In case of a relatively moving object this error increases. For the simulated TV-sequence a separation between relatively moving area and background has been achieved if the relative displacement exceeds a threshold of 2.4 pel per frame.
KeywordsExpense Cross Correlation Agram
Unable to display preview. Download preview PDF.
- /1/.Bohner, M. “Objektermittlung in Bildfolgen” in Erfassung und maschinelle Verarbeitung von Bilddaten ed. by H. Kazmierczak Springer Verlag, Wien 1980Google Scholar
- /2/.Huang, T. S., “Motion Estimation” in Tsai, R. Y, Image Sequence Analysis ed. by T. S. Huang Springer Verlag, Berlin 1981Google Scholar
- /3/.Beyer, S. “Displacement Measurement with Fast Video Correlators” in Proceedings of the Picture Coding Symposium Montreal (Canada) 1981, pp. 56–57Google Scholar
- /4/.Castleman, K. R. Digital Image Processing Prentice Hall, Inc., Englewood Cliffs 1979Google Scholar