A Fast Non-Local Means Image Denoising Algorithm

Mobile devices, such as smart phones, are constrained to the image/video processing capacity due to their limited battery life and computational power. In order to cope with these bottlenecks, in this paper, we propose a novel fast non-local means image denoising technique for mobile image processing. The major steps of the proposed technique includes: we reduce the computation of pixels utilizing the symmetry of pixel weights; in the sampling optimization step, certain pixels are selected and their weights are computed by the weighted formula while the weights of the unselected pixels are computed by the bilinear interpolation method; we optimize the codes of the program with NEON instructions in the assembly level. The experiments show that: the image denoising time of the proposed technique is reduced effectively compared to the traditional non-local means image denoising technique with little image quality loss, which satisfies the real-time requirement on mobile devices.