Fixed-Coefficient Iterative Bilateral Filters for Graph-Based Image Processing

摘要

We present a graph-based image processing algorithm using fast iterative bilateral filters. The computation of bilateral filters is accelerated with fixation of the coefficients during iterations and their approximate decomposition further speeds up the computation. We show that this fixed-coefficient iterative bilateral filter is an alternative solver for optimization problems in graph-based data analyses and apply its fast algorithm to graph-based image processing tasks. Performance of the present algorithm is demonstrated with experiments of contrast enhancement and smoothing of images using cross bilateral filters, in addition to semi-supervised image segmentation and colorization of monochromatic images.