Performance and Complexity Comparison between H.265/HEVC and H.264

HEVC Test Model under Consideration (TMuC) serves as a starting point for definition of the new video coding standard. The current version is HM-4.1-dev.

An article "Towards high efficiency video coding: Subjective evaluation of potential coding technologies" pointed out:

"The TMuC has similarities to the H.264/MPEG-4 AVC standard, including block-based intra / inter prediction, block transform and entropy coding. New features include increased prediction flexibility, more sophisticated interpolation filters, a wider range of block sizes and new entropy coding schemes. Twice the compression efficiency of H.264/MPEG-4 AVC is expected to be achieved, at the expense of a considerable increase in computational complexity. The performance of the coding algorithm resulting from this integration step will be analyzed by means of formal subjective quality assessment in a next subjective test campaign."

The article "High Performance, Low Complexity Video Coding and the Emerging HEVC Standard" describes a low complexity video codec with high coding efficiency. It was proposed to the high efficiency video coding (HEVC) standardization effort of moving picture experts group and video coding experts group, and has been partially adopted into the initial HEVC test model under consideration design. The proposal utilizes a quadtree-based coding structure with support for macroblocks of size 64 × 64, 32 × 32, and 16 × 16 pixels.

• Entropy coding is performed using a low complexity variable length coding scheme with improved context adaptation compared to the context adaptive variable length coding design in H.264/AVC. 
• The proposal's interpolation and deblocking filter designs improve coding efficiency, yet have low complexity. 

Dan Grois also summarized

• Average interpolation complexity is less than that of H.264/AVC due to the 1-D directional interpolation filters. 
• Memory bandwidth for motion compensation is lower than H.264/AVC. This is because motion partitions smaller than 8 × 8 are not used. 
• Interpolation complexity for B pictures is significantly lower in the proposal than in H.264/AVC, as the SKIP and DIRECT modes use integer motion vectors. Using motion vectors with integer pixel accuracy implies that reconstruction of a SKIP/DIRECT block can be done by copying pixels from one location in memory to another location, instead of applying the 6-tap interpolation filter. 
• The deblocking filter has significantly lower computational complexity. This is mainly because a much lower number of block edges are filtered, and the filter has simpler logic for enabling/disabling the filter on an edge. In addition, filtering can be performed in parallel first for each of the vertical edges and then for each of the horizontal edges, which is not possible for H.264/AVC. 
• VLC (de)coding of coefficients is simpler than H.264/AVC CAVLC coefficient (de)coding. In particular, CAVLC relies on decoding a large number of syntax elements (e.g., coeff−token, trailing−ones−sign−flag, level−prefix, level−suffix, total−zeros, and run−before). Also, the VLC decoding uses significantly fewer syntax elements, and requires less conditional branches (run/level/sign in run mode, and level/sign in level mode). 

Samuelsson at el. proposed a method to for "Simplification of Intra Prediction Mode" for HEVC. The result shows that this processing is adaptive to the content of the video, and results in a reduction in the computational complexity of the prediction process by approximately 50%.