Bit-depth Scalable Coding for High Dynamic Range Video
Conventional video sources are represented in 8bits per pixel (bpp) component, while many professional applications require video sources in higher bit-depths, i.e. 10, 12, 14, etc. bits per pixel component. Given a higher bit-depth input video, the compressed bit stream should consist of two (or more) layers, where the base layer contains 8 bpp content, which can be displayed on conventional 8 bpp consumer display devices; while the enhancement layer carries higher bit-depth content, which can fulfill professional application requirements. This multi-layer bit stream can be easily achieved by simulcast; however, the compression efficiency is very low.
Background & Objective: The key issue in bit-depth scalable coding is to predict higher bit-depth pixels (e.g. 10 bpp) from base layer (e.g. 8 bpp) pixels, which is referred as inter-layer prediction. The inter-layer prediction can be considered as an inverse tone mapping procedure, whose performance is strongly related to the forward tone mapping method used to convert pixels from higher to lower bit-depth. However, the tone mapping method is not always known and it may have variations in many ways. For example, different tone mapping methods, such as linear or non-linear, can be used for various applications; it may also vary in different frames, scenes, or in different regions within one picture; it may vary in different color channel as well; and the dynamic range can also be big or small.
Technical Discussion: In order to cope with bit-depth scalability, several contributions have recently been proposed in JVT standard meetings. With extensive study of the prior arts, we developed a novel inter-layer prediction scheme and proposed it to JVT standard group. Our proposed solution is localized (block-based), which is more efficient for coding videos generated by local tone maps in different scenes, pictures, and regions of interest. The proposed solution is also independent to each color channel; thus, it provides the flexibility and efficiency to compress videos in various color spaces. Further more, our approach is suitable for a wide variety of tone mapping techniques, and can cover a wide bit-depth dynamic range. Experimental results show that that our proposed scheme achieves significant coding gain over other localized (block-based) approaches. Coding gains over global approaches on locally tone mapped videos are expected.
Contacts:
Shan Liu
Anthony Vetro
Publications:
Shan Liu, Woo-Shik Kim, Anthony Vetro, "Bit-depth Scalable Coding for High Dynamic Range Video", SPIE Conference on Visual Communications and Image Processing (VCIP), January 2008 (TR2007-078)
Technology Area: Multimedia
Modification Date: July 5, 2007
