Repetitive and morphological padding for object-based video coding

This paper introduces a new technique in block-based motion estimation and motion compensation for object-based video coding. Out of many approaches for object-based coding, hybrid algorithms that take advantages of the traditional motion compensated block-based DCT approach are noted for their maturity and simplicity. Indeed, it has become the main feature of the upcoming international video standard MPEG-4. In this type of hybrid algorithm, the boundary of the objects calls for special attention. The discontinuity at object boundary could be very costly if not handled properly. We discovered that efficient motion estimation and compensation for the object boundary is most critical to the performance of object-based video coding. Our studies show that padding, a process of boundary value extension, is essential in motion estimation/compensation of arbitrarily shaped objects. We have designed novel padding algorithms and report our preliminary results in this paper