Background: High-Efficiency Video Coding (HEVC) is a recent video compression
standard. It provides better compression performance compared to its predecessor, H.264/AVC.
However, the computational complexity of the HEVC encoder is much higher than that of
H.264/AVC encoder. This makes HEVC less attractive to be used in real-time applications and in
devices with limited resources (e.g., low memory, low processing power, etc.). The increased computational
complexity of HEVC is partly due to its use of a variable size Transform Unit (TU) selection
algorithm which successively performs transform operations using transform units of different
sizes before selecting the optimal transform unit size. In this paper, a fast transform unit size selection
method is proposed to reduce the computational complexity of an HEVC encoder.
Methods: Bayesian decision theory is used to predict the size of the TU during encoding. This is
done by exploiting the TU size decisions at a previous temporal level and by modeling the relationship
between the TU size and the Rate-Distortion (RD) cost values.
Results: Simulation results show that the proposed method achieves a reduction of the encoding
time of the latest HEVC encoder by 16.21% on average without incurring any noticeable compromise
on its compression efficiency. The algorithm also reduces the number of transform operations
by 44.98% on average.
Conclusion: In this paper, a novel fast TU size selection scheme for HEVC is proposed. The proposed
technique outperforms both the latest HEVC reference software, HM 16.0, as well as other
state-of-the-art techniques in terms of time-complexity. The compression performance of the proposed
technique is comparable to that of HM 16.0.