Adaptive Quantization-Based HDR Video Coding with HEVC Main 10 Profile

Perceptual quantization (PQ) for HDR video coding has good perceptual uniformity in the luminance range with the modest bit-depth. However, the dynamic range of HDR videos is not well-used in PQ according to contents, especially for chroma channels. Thus, there exist the wasted dynamic ranges in PQ which cause detail loss and color distortions. In this paper, we propose adaptive quantization-based HDR video coding with HEVC Main 10 Profile. We perform adaptive mapping for 16bit to 10bit conversion based on cumulative distribution function (CDF), i.e. adaptive quantization, instead of linear mapping. First, we obtain CDF of each video frame using histogram analysis. Then, we conduct adaptive quantization based on CDF according to video contents, and generate metadata for inverse adaptive conversion in decoder, i.e. 10bit to 16bit conversion. Experimental results demonstrate that the proposed method reconstructs high-quality HDR videos and achieves a significant improvement in performance in terms of visual assessment and quantitative measurement.

I. Introduction

High dynamic range (HDR) videos remarkably enhance the visual experience compared to low dynamic range (LDR) ones. However, they need a huge amount of the storage space and transmission bandwidth because of the high bit-depth for represent the dynamic range [5] [6]. High Efficient Video Coding (HEVC) is the most recent video coding standard which reduces nearly 50% bitrate compared with H.264/AVC [7]. In 2015, MPEG released an anchor framework (Anchor) [1] for HDR/Wide Color Gamut (WCG) video coding based on HEVC Main 10 Profile. In recent years, researchers have proposed several solutions to HDR video coding. Mantiuk et al. [8] [9] proposed an HDR video coding scheme in the hybrid luminance-frequency space. They considered the spatial frequency sensitivity characteristic of Human Visual System (HVS) and calculated the difference between neighboring pixels, which saved low frequency and edge information in DCT block and edge block separately. Zhang et al. [10] considered the luminance masking effect of HVS and proposed the intensity difference quantization (IDQ) profile for quantizing code residual with different quantization steps according to the average luminance of one block. Motra et al. [11] proposed an adaptive LogLuv transform method for HDR video coding. In this method, they transformed the RGB color space into the integer LogLuv color space, and minimized the quantization error in the mapping stage. However, Anchor [1] causes much color distortion and detail loss in the reconstruction result. As shown in Fig. 1, Anchor [1] causes color distortions in the woman's arm compared with the original HDR image. We observe that the pixel values in chroma channels are generally concentrated on a relatively narrow dynamic range and thus the dynamic range of HDR video is not well-used as shown in Fig. 2. This narrow range would be much narrower after being linearly quantized to a lower bit depth, which leads to color distortion and detail loss in chroma channels. Fig. 1.

Original HDR image and its reconstructed HDR result by anchor [1].

Fig. 2

Histograms of Y, U, V channels of a frame in HDR video market.