The Messy-Niche algorithm used for image super-resolution

Two problems will be met with while directly using simple genetic algorithm to solve image interpolation: one is how many bits are needed for difference-chromosome coding. Setting it too large will increase implementation spending while making it too small will evolve into local minimum. The other is premature, if some individuals are most dominant in current population, they will emphasize their superiority in future evolution, so the diversity is restrained and evolution will be plunged into local minimum. This paper investigates messy algorithm and niche algorithm which can easily remedy the two shortcomings above. For the purpose of extracting an ultimate algorithm, messy algorithm and niche algorithm are combined. Experiments on the application of image super-resolution demonstrate the algorithm proposed has better performance than messy algorithm or niche algorithm used alone.