Shape-Adaptive Discrete Cosine Transform (SA-DCT) has been proposed for coding arbitrarily shaped objects. SA-DCT was used in lieu of DCT, and suitable modifications were made to the MPEG- 2 codec to provide access to objects comprising an image. It was established that this additional functionality was not obtained at the cost of compression efficiency. SA-DCT along with contour coding, using chain-difference method, was verified to be an efficient method for coding arbitrairly shaped objects by performing experiments with standard test images.

This paper has previously appeared as a Technical Report at Philips Research.