We propose a general solution to the problem of decoding high dynamic range (HDR) information stored as a supplement to low dynamic range (LDR) images or video. Each LDR frame is paired with a lower resolution HDR version, and these are compressed separately using any of the existing methods appropriate to the task. On decode, the low-resolution HDR image is upsampled to match the resolution of the LDR version, and the high frequencies are transferred from the LDR to the HDR frame. The recovery process places no constraints on the color space or tone-mapping of the backwards-compatible LDR content, and is thus ideally suited to applications such as DVD movies that target legacy equipment while building in forward-compatibility with emerging HDR systems. A fast and simple recovery algorithm is demonstrated, followed by a more sophisticated and accurate technique. Examples are shown on computer-generated video frames as well as HDR captured video.
展开▼
