This paper presents the results of research in the use of genetic programming for image compression. While genetic programming has been examined for image compression, we describe here a new method that divides the image into variable-sized blocks using an adaptive quad-tree subdivision algorithm. We also examine the choices of function definitions for the genetic programs, and their effect. This work shows that there may be more effective solutions for image compression that use evolutionary approaches than some currently known methods of image compression such as vector quantization. The genetic programming system itself is also a unique design, using Microsoft's C#.NET language and environment. This allows for symbolic manipulation of the programs (chromosomes); followed by compilation to memory, and then execution as machine code. This approach combines ease of manipulation with efficient execution.
展开▼
