Modified bilateral filtering algorithm with gradient weight factor

Motivated by the traditional bilateral filtering, apart from considering geometric closeness and photometric similarity, we introduce the gradient factor to form a new filter kernel. Gradient reflects to the trend of edge and texture directly. By introducing the gradient weight factor, the proposed filtering algorithm overcomes the isotropy of the traditional Gaussian function, and has the directional filtering effect. Expanding from the direction of the gradient to both sides, pixels approach the trend of edges gradually. When replaces the current pixel by a weighted average of its neighbors, the proposed algorithm assigns a greater weight to the pixels that approach the trend of edges, in order to enhance the contrast of the edges. Experimental results show that, through the introduction of the gradient weight factor, the proposed method can not only restrain noise, but also keep much more weak edges and details of the input images.