Structure Preserving Semantic Texture Filtering

In structure preserving texture filtering, structural edge based features are more useful than textural features as they are invariant for both regular and varying scale irregular textures. In this work, we proposed a novel structure preserving texture filtering technique by detecting structural edges of the image. The proposed technique first exploits Jensen Shannon Divergence to detect structural edges of the image considering semantic information. Then with help of the detected structural edges, it defines a dynamic window of the median filter for filtering the input image. The experimental results on a variety of images show the potentiality of the proposed technique.