Adversarial and Adaptive Tone Mapping Operator for High Dynamic Range Images

Abstract

This work addresses tone mapping, a common approach to convert high dynamic range (HDR) images into low dynamic range (LDR) images. We approach this problem by using adaptive tone mapping. We propose to deploy a conditional generative adversarial networks to build an adversarial and adaptive tone mapping operator (adTMO) that converts HDR into LDR images. We use an objective quality metric called the Tone Mapped Image Quality Index (TMQI) to evaluate our adTMO. Trained with 256*256 images, adTMO is able to generate 256*256 and high-resolution 1024*2048 LDR images. Given 1024*2048 HDR images, TMQI of the generated LDR images reaches the value of 0.90, which outperforms all other contemporary tone mapping operators.