An efficient Lagrangian multiplier selection method based on temporal dependency for rate-distortion optimization in H.265/HEVC

Rate-distortion optimization is a crucial instrument to achieve the best performance of information lossy video coding based on rate-distortion theory, and it is generally realized by Lagrange optimization using a Lagrangian multiplier. An efficient Lagrangian multiplier selection method considering temporal dependency (LMS-TD for short) is introduced in this paper for the random-access (RA) configuration in H.265/HEVC (High Efficiency Video Coding). The temporal relationship of distortion and bit-rate is analyzed for videos coded using an H.265/HEVC compliant coder (HM-14.0) with the RA configuration, based on which an optimized set of Lagrangian multipliers is determined. The LMS-TD method achieves an average BD-rate (Bjontegaard bit-rate) reduction of 4.59%. Picture quality evaluation was also performed using an objective perceptual video quality measurement technique, i.e., the Structural Similarity Index (SSIM), where the LMS-TD method demonstrated superior performance in video quality.