Dual-statistic white balance guided 3DLUT for image enhancement

The 3D Lookup Table (3DLUT)-based methods are increasing popularity due to their satisfying and stable performance in achieving automatic and adaptive real time image enhancement. In this paper, we present a new solution Dual-statistic white balance guided 3DLUT to compensate for the color deviation of the 3DLUT due to its own structure. The color correction based 3D Lookup Table is achieved on the RGB three color channels attributes to the effective extraction of color features from the design of Dual-statistic white balance Encoder. Specially, we compute the affinity matrix to match color and content features from the design of the Dual affinity matrix Module to achieve more realistic and satisfying enhancements based on the improved fidelity and consistency of color information. Our approach can be applied in image retouching and tone mapping tasks with fairly good generality. The performance in both theoretical analysis and comparative experiments manifests that the solution we proposed is effective and confirmed.