Deep Controllable Backlight Dimming for HDR Displays

High dynamic range (HDR) displays with dual-panels are one type of that can provide HDR content. These composed a white backlight panel and colour LCD panel. Local dimming algorithms used to control the in order reproduce content high contrast at fidelity. However, existing local usually process low (LDR) images, which not suitable for processing images. In addition, these methods use hand-crafted features estimate values, may be many kind this work, novel deep learning based method is proposed rendering images on dual-panel displays. The uses Convolutional Neural Network (CNN) directly predict using as input image displayed. model designed trained via controllable power parameter allows user trade off between quality. evaluated against seven other test set 105 variety quantitative quality metrics. Results demonstrate improved display better consumption when compared best alternatives.