This paper presents a novel analysis and synthesis approach to image compression. In the encoding stage, an analysis scheme is used to selectively encode visually important features of an image to achieve computational and compression efficiencies, and in the decoding stage, a synthesis method is employed to reconstruct the full image from the selectively encoded image features. By integrating established image coding technologies and the latest development in computer graphics, especially texture synthesis and image inpainting, our method opens up a promising new direction for developing novel solutions to image compression and related problems. Preliminary experimental results are presented to demonstrate the practicability of this innovative image compression method.
[1]
Guoping Qiu,et al.
Fast algorithm for constructing image identification
,
1995,
Defense, Security, and Sensing.
[2]
Tony F. Chan,et al.
Mathematical Models for Local Nontexture Inpaintings
,
2002,
SIAM J. Appl. Math..
[3]
Sung Yong Shin,et al.
On pixel-based texture synthesis by non-parametric sampling
,
2006,
Comput. Graph..
[4]
Image coding based on visual vector quantization
,
1995
.