Determination of bilateral filter coefficients based on particle swarm optimization

Image restoration has been considered as the front end of the image processing pipeline. Its performance would critically affect other subsequent operations such as object detection and recognition. Color image enhancement, with simultaneous noise reduction and contrast preservation, is one of the difficult problems to be tackled. Among the possible approaches available, bilateral filtering (BF) is known for its specially designed features that address the enhancement problem. However, it still remains non-trivial to determine proper filter coefficient settings. This challenge is here cast as an optimization problem on filter parameters. In particular, the particle swarm optimization (PSO) algorithm is adopted for its solution quality and implementation simplicity. Filter coefficients are coded as particles and solutions are iteratively improved against the signal-to-noise criterion. Experiments are conducted on publicly available color images. Results are used for recommending a set of feasible filter coefficient values.