Efficient lossless bit depth scalable coding for HDR images

This report proposes two layered bit depth scalable coding methods for high dynamic range (HDR) images expressed in floating point data format. From the base layer bit stream, low dynamic range (LDR) images are decoded. They are tone mapped appropriately for human eye sensitivity, and shortened to a standard bit depth, e.g. 8 [bit]. From the enhance layer bit stream, HDR images are decoded. However the bit depth of this layer has been huge in the existing method. To reduce it, we divide the tone mapping into a reversible logarithmic mapping and its compensation. It was confirmed that the proposed methods significantly reduce the bit depth of the enhance layer, even though the compensation slightly increases coding noise.