Multi-Exposure Image Compression Considering Rate-Distortion Optimization in Rendered High Dynamic Range Image

Abstract

High-dynamic range (HDR) images, which provide realistic visual perception, have attracted considerable attention in various applications. A simple method for obtaining HDR images is to fuse multiple low-dynamic range (LDR) images captured under varying exposures. We proposed a method of efficiently encoding multi-exposure images using multi-view High Efficiency Video Coding (MV-HEVC) along with intensity mapping function (IMF) to create high-quality HDR images. The inter-view prediction in MV-HEVC eliminates redundancy between multi-exposure images. To achieve high-efficiency MV-HEVC-based multi-exposure image coding, two modifications to the rate-distortion optimization (RDO) process performed during encoding were proposed. First, the distortion in the rendered HDR image is used to replace the original distortion, which depends on the LDR image. Second, to balance distortion and rate in RDO, a new Lagrange multiplier is derived from the derivative of the new distortion with respect to the rate. A model of the distortion in the HDR domain was constructed and expressed as a function of the distortion in the LDR domain. The proposed scheme utilizes the conventional 8-bit codec to compress multi-exposure LDR images and generate an HDR image in the decoder. The experimental results indicate that the proposed technique outperforms HDR image coding schemes in which HEVC Range Extension and JPEG-XT coding standards are used.