My interest in video coding has been mainly in the joint analysis of transform coding and quantization.

My recent work with Guan-Ming Su (Dolby Lab) and Min Wu (UMD) analyzed how reshaping/baseband quantization affects the coding efficiency. Our theoretical analysis shows that for the best coding efficiency, video information reduction is better incurred at the quantizer in the video codec than at the baseband quantizer/in the reshaping stage. Our work gives a quantitative explanation for the balance between baseband quantizer and codec quantizer, which can be helpful for practitioners’ decision-making.

The theoretical result on triple quantizers with interleaved transforms (see page 7 of the link, or the supplementary file on IEEEXplore) is also brand new, and we are not aware of similar results in the literature on information theory and video coding.

The source code for this work can be downloaded at the bottom of this page or from IEEEXplore. A poster for this work can be downloaded here. A video explanation of this work is shown as follows.