Local look-up table upsampling for accelerating image processing

Abstract The resolution of cameras is increasing, and speedup various image processing required to accompany this increase. A simple way acceleration the at low then upsampling result. Moreover, when we can use an additional high-resolution as guidance formation for upsampling, upsample results more accurately. We propose approach accelerate by downsampling joint upsampling. This paper utilizes per-pixel look-up tables (LUTs), named local LUT, which are given a low-resolution input output pair. Subsequently, LUT. generate only referring its In our experimental results, evaluated proposed method on several filters applications: iterative bilateral filtering, $$\ell _0$$ ℓ 0 smoothing, Laplacian inpainting, haze removing. accelerates with sufficient approximation accuracy, outperforms conventional approaches in trade-off between accuracy efficiency. Our code available https://fukushimalab.github.io/LLF/ .