DCT and wavelet based representations of arbitrarily shaped image segments

This paper describes two efficient algorithms for representing arbitrarily shaped objects. The first representation, called cosine filling, is a DCT-based approach for obtaining a compact description of objects, making it possible to achieve high compression. A wavelet-based approach is also proposed. This representation not only provides compact description, but also facilitates smooth extension, which can be useful in extrapolation and interpolation of objects.