Proposal of energy efficient transform cores to be used in the High Efficiency Video Coding (HEVC)

What is it about?

In this paper, we report on the effects of reducing the bit length of the elements of the HEVC transforms on video compression performance. This is significant because it has been demonstrated that the proposed cores yield an efficient video compression. They also permit a substantial gain in hardware resources, in particular the processing speed and the energy consumption. The paper should be of interest to readers in the areas of video/image compression, telecommunications and FPGA-based implementation. HEVC is a highly efficient video compression standard, but its complexity is a serious drawback. Many researchers are dealing with this issue by presenting different transforms with reduced complexity. This work considers the problem from a different perspective. The HEVC transforms are 8 bits coded and to be compliant with the HEVC requirements, the outputs should never exceed 16 bits. This condition is achieved through the truncation of the intermediate results. This is, in our opinion a waste of resources. Therefore, our idea is to reduce the length of the transforms elements from 8 to 6. The resulting transform kernels, namely the integer DCT of size 4x4, 8x8, 16x16, 32x32 and the integer DST 4x4 were evaluated for their basic mathematical properties requirements, their computational complexity, their FPGA-based implementation metrics and their compression efficiency when embedded in the HEVC algorithm. The results are very interesting and competitive with the recently published works.

Featured Image

Why is it important?

HEVC is a highly efficient video compression standard, but its complexity is a serious drawback. Many researchers are dealing with this issue by presenting different transforms with reduced complexity. This work considers the problem from a different perspective and proposes different transform kernels that are: 1) More precise 2) Ensuring the best trade-off between video quality and the bit rate (size of the bit stream). 3 ) Providing a relatively easy implementation on FPGA devices. 4) Energy efficient which is the most important feature in the new multimedia tools

Perspectives