Adaptive boundary filtering

In this paper, we present a new adaptive boundary filtering which is an improvement of the bilateral filtering algorithm. Bilateral filtering is an algorithm that is used for smoothing and preserving the edges in an image. This algorithm has proper capability when using with normal surface that have less number of regions. For texture area, many complicated regions are packed together e.g., hair and wood. Bilateral filtering filters this area as the unsmooth surface because there are many tiny regions located in the same filtering boundary. Therefore, textures are eliminated and blurred. We solve this problem by applying the new type of boundary to the filtering. The pixel aggregation, an image segmentation algorithm, is used for searching this new type of boundary. In normal surface, this filtering boundary is similar to bilateral filtering then it filter the image as bilateral filtering does. For the texture area, this boundary is auto adapted itself to cover only relevant regions. Therefore, it excludes the unrelated regions that cause problem to the texture then the details in texture are preserved. Our algorithm is tested with various images. The result shows that adaptive boundary filtering can smooth related area and then textures are preserved.