Makishi et al. (2017) proposed a method that estimates the parameters for a bilateral filter, based on the distribution distance of images. This study proposes an improved version of the Makishi et al. (2017) method, which includes three new contributions. Firstly, our improved method was verified with two image sizes and two noise distributions. Secondly, five distribution distance functions were examined for the local search of parameter sets, where SKLD was found to be slightly superior to HD in terms of image quality and execution time. Finally, the parallelization of local search was examined for faster parameter estimation with recent multi-core processors. With a 2.8GHz Xeon processor of 10 cores, the parameter estimation times were reduced to 36ms (256x256 pixel) and 1.03 s (1024x1024 pixel).
J-STAGEがリニューアルされました！https://www.jstage.jst.go.jp/browse/-char/ja/
