Fast texture‐preserving bilateral filtering

A new method for fast bilateral filtering with texture‐preserving properties is presented. In general, the texture area is composed of several tiny regions that have almost similar intensity, and therefore bilateral filtering combines similar gray scales of the texture area to form a smooth region. Adaptive boundary filtering solves this problem by using image segmentation to define a new weighting boundary at each pixel filtering. With the new type of boundary, the output image is smoothed while preserving both edges and textures details. The drawback of this algorithm is that it is time consuming. In this paper, a histogram‐based filtering is proposed to improve the speed of this algorithm. However, as texture details are blurred when a small number of bins are used, it is unable to increase speed by using a small number of bins. Deciles‐based algorithm is therefore applied to overcome this limitation, as it can increase the speed while preserving the texture details. The new algorithm, named fast adaptive boundary filtering, increases the speed by more than 40% when compared with adaptive boundary filtering at a boundary threshold T = 40. The output image of new algorithm shows that it is similar to the output of adaptive boundary filtering. Moreover, the speed of new algorithm is compared with the normal 256‐bin histogram‐based bilateral filtering. From the experimental results, it is seen that the new algorithm has higher speed when it is processed with a small to moderate radius window. In addition, a better output image can be obtained from the new algorithm as it preserves both edges and textures. © 2015 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.