Domain-Aware Adaptive Logarithmic Transformation

Tone mapping (TM) aims to display high dynamic range scenes on media with limited visual information reproduction. Logarithmic transformation is a widely used preprocessing method in TM algorithms. However, the conventional logarithmic does not take difference image properties into account, nor it consider tone algorithms, which are designed based luminance or gradient-domain features. There will be problems such as oversaturation and loss of details. Based analysis existing methods, this paper proposes domain-aware adaptive AdaLogT for We introduce parameter p construct different objective functions domains algorithms determine optimal values adaptively. Specifically, luminance-domain we use exposure histogram features function; while texture-aware exponential mean local variance (EMLV) build function. Finally, propose joint deep-neural-network-based The experimental results show that novel endows each domain algorithm wider scene adaptability improves performance terms effects evaluations, subjective index scores quality improved by 6.04% 5.90% average