Multiple non-orthogonal bases representations for images

Conventional image compression techniques usually involve use of a single transform in conjunction with other techniques like vector quantization etc. In this paper, the concept of using more than one transform, i.e., using multiple non-orthogonal bases functions representation, for still image compression, is presented. Sample results show that this approach used in conjunction with an efficient vector quantization technique - an adaptive energy based split vector quantization technique, gives improved reconstruction quality of images for the same bit rate compared to existing single transform methods.