Method and apparatus for coding an image object of arbitrary shape
Liou, Ming L.
|Source||US Patent , 6,718,066 B1, 2004|
|Summary||At the internet age, transmittal of video signal in digital form is very common. Among various video compression standards or processes established, Discrete cosine transform (DCT)- based block transform coding is one of the frequently used coding techniques for compression of still image. This method provides a compromise between energy packing ability and computation complexity. However, the existing shape-adaptive DCT (SA-DCT) method suffers from its non-regular structure and the non-existence of efficient implementation methods. Its hardware implementation requirement can make fabrication very costly. This invention offers a new method and apparatus based on which the arbitrarily-shaped transform coding can be accomplished using the regular N-point DCT/IDCT. The existing fast algorithms and/or available chipsets could be thus utilized directly. The new method consistently provides several dB improvement over the LPE padding that is recommended in MPEG-4, at just a minimal increase in computation requirement at the encoder side only. It has low computational complexity, slightly better encoding performance, and no requirement for depadding process at the decoder side. Last but not least, this method can be easily integrated into any object-based video/image control system that supports arbitrarily-shaped coding functionally.|
|Rights||This patent is also available at HKUST Technology Transfer Center Patent Search at <http://patentsearch.ttc.ust.hk/ListPatentsDetail.asp?id=148>|
Files in this item: