High Dynamic Range Images Coding: Embedded and Multiple Description

The aim of this work is to highlight and discuss a new paradigm for representing high-dynamic range (HDR) images that can be used for both its coding and describing its multimedia content. In particular, the new approach defines a new representation domain that, conversely from the classical compressed one, enables to identify and exploit content metadata. Information related to content are used here to control both the encoding and the decoding process and are directly embedded in the compressed data stream. Firstly, thanks to the proposed solution, the content description can be quickly accessed without the need of fully decoding the compressed stream. This fact ensures a significant improvement in the performance of search and retrieval systems, such as for semantic browsing of image databases. Then, other potential benefits can be envisaged especially in the field of management and distribution of multimedia content, because the direct embedding of content metadata preserves the consistency between content stream and content description without the need of other external frameworks, such as MPEG-21. The paradigm proposed here may also be shifted to Multiple description coding, where different representations of the HDR image can be generated accordingly to its content. The advantages provided by the new proposed method are visible at different levels, i.e. when evaluating the redundancy reduction. Moreover, the descriptors extracted from the compressed datastream could be actively used in complex applications, such as fast retrieval of similar images from huge databases.