PSO-Based Fast Image Analogies With Its Applications

Analogy is a basic reasoning process, one that we as humans employ quite commonly, and often unconsciously, to solve problems, provide explanations, and make predictions. Image analogy is based on learning. It is that the computer analyzes and learns the relation of the two images in source images, and then applies the relation to the two target images by using the analogy learning from the people.The analogies algorithm is proposed by Hertzmann and so on who get the inspiration from the process of humanity discursion. This algorithm solved the question of learning "art filtered mirror" from the digital photorealistic rendering. And the algorithm does perfect job in traditional image filters, improved texture synthesis, texture transfer, texture-by-numbers and so on. The idea of analogies algorithm illumine spacious experts and broaden the application of analogies algorithm. However, the low efficiency of closest-matching pixel affects the efficiency of analogies algorithm.This paper releases fast image analogies algorithm based on particle swarm optimization (PSO). PSO is proposed by Kenndey and Eberhart in 1995, and it is a new evolution algorithm based on swarm intelligence. PSO being iterations random algorithm does better in fast constringency and finding the optimization solution with biggish probability.This paper uses PSO in image analogies instead of using approximate-nearest-neighbor search (ANN). The algorithm speed is improved as much times as former speed and the algorithm outcome is the same to former outcome. The new algorithm is better in having small parameter, being easy to carry out, double-quick constringency and accommodating parameter easily.The fast image analogies algorithm based on particle swarm optimization will have broad application in traditional image filters, improved texture synthesis, texture transfer, texture-by-numbers and so on.