Accelerated Multimodal Images Joint Non-local Filtering

In this paper, we present a novel non-local averaging based multimodal image filtering method. In the method the mean average for filtering the noisy image is computed nonlocally and weights for average is jointly guided by the reference image taken from another imaging device. The experiment results shows that the proposed method is out performance than the state of the art joint bilateral filter, especially in the case of the periodic texture is involved in the images. Further more, we accelerate the method by constructing integral histograms, speed it up by factor of about 2 order compared with the brute-force method. In addition, the algorithm is independent of the size of the filter, and easy to be implemented on parallel system.