Efficient burst image compression using H.265/HEVC

New imaging use cases are emerging as more powerful camera hardware is entering consumer markets. One family of such use cases is based on capturing multiple pictures instead of just one when taking a photograph. That kind of a camera operation allows e. g. selecting the most successful shot from a sequence of images, showing what happened right before or after the shot was taken or combining the shots by computational means to improve either visible characteristics of the picture ( such as dynamic range or focus) or the artistic aspects of the photo ( e. g. by superimposing pictures on top of each other). Considering that photographic images are typically of high resolution and quality and the fact that these kind of image bursts can consist of at least tens of individual pictures, an efficient compression algorithm is desired. However, traditional video coding approaches fail to provide the random access properties these use cases require to achieve near-instantaneous access to the pictures in the coded sequence. That feature is critical to allow users to browse the pictures in an arbitrary order or imaging algorithms to extract desired pictures from the sequence quickly. This paper proposes coding structures that provide such random access properties while achieving coding efficiency superior to existing image coders. The results indicate that using HEVC video codec with a single reference picture fixed for the whole sequence can achieve nearly as good compression as traditional IPPP coding structures. It is also shown that the selection of the reference frame can further improve the coding efficiency.