Fast Context-Adaptive Bit-Depth Enhancement via Linear Interpolation

To fill in the huge gap between the rapid development of modern display devices with high dynamic range (HDR) and the mainstream media source with a low bit depth of 8, bit-depth enhancement (BDE) that reconstructs high bit-depth images from low bit-depth ones emerges as an important problem. Early works generally introduce the annoying artifacts and then plenty of context-aware algorithms are proposed recently but at the cost of high computational complexity. In this paper, we propose a fast yet efficient context-adaptive algorithm via linear interpolation. For gradual color transition areas, the interpolation is performed along the direction bisecting the self-defined virtual effective boundaries, while for local peak/bottom areas, the interpolation is carried out based on the minimum distance to intensity steps caused by quantization. By deliberately designing the interpolation direction and endpoints for different types of local contexts, the proposed algorithm can well reconstruct natural images with low computational complexity. The experimental results show that the proposed algorithm outperforms the state-of-the-art algorithms on average in terms of both objective and subjective evaluations.