Perceptual-based coding mode decision

Abstract

The framework of rate-distortion optimization (RDO) has been widely adopted for video coding to achieve a good trade-off between bit-rate and distortion. However, objective distortion metrics such as mean square error traditionally used in this framework are poorly correlated with perceptual video quality. To address this issue, we incorporate the structural similarity index as a quality metric into the framework and develop a predictive Lagrange multiplier selection technique to resolve the chicken-and-egg dilemma of perceptual-based RDO. The resulting perceptual-based RDO is then applied to H.264 intra mode decision as an illustration of the application of the proposed technique. Given a perceptual quality level, 5%-10% bit rate reduction over the JM reference software of H.264 is achieved. Subjective evaluation further confirms that, at the same bit-rate, the proposed perceptual RDO preserves image details and prevents block artifact better than the traditional RDO.

A novel low complexity Local Hybrid Pseudo-SSIM-SATD distortion metric towards perceptual rate control

“…The scaling is performed using linear equations to ensure processor friendly operations. Compared to [8], LHPSS can operate locally without using logarithmic functions and without re-quantising to produce a temporal relative distortion scale. As LHPSS operates in the distortion metric space of SATD it satisfies the triangle equality rule ( ).…”

“…As LHPSS operates in the distortion metric space of SATD it satisfies the triangle equality rule ( ). Therefore, in [8] the perceptual model must be actively updated on a key frame basis as its relative nature limits the scope to the GOP level or when there is high activity. Hence, in [8] it does not adapt to the local conditions like an STDM.…”

“…Therefore, in [8] the perceptual model must be actively updated on a key frame basis as its relative nature limits the scope to the GOP level or when there is high activity. Hence, in [8] it does not adapt to the local conditions like an STDM. LHPSS distortion metric addresses these issues related to [8] by providing a general solution based upon modelling the UBR by block size [2].…”

“…The efforts of adapting SSIM to operate as a pseudodistortion metric have been achieved in [8] by use of logarithmic functions, however this approach is limited to the Group of Pictures (GOP) level. Furthermore, it does not meet the goals of low complexity and variability [3].…”

“…Finally, the results of the implemented LHPSS, in terms of It is shown in [5] that SSIM must be non-negative, symmetrical and fulfil the triangle equality rule ( ). These first two conditions are met by presenting SSIM in the form of (1-SSIM), a method adopted in [8] and explained in [5]. To support the triangle equality rule ( ), the findings in [2] of a universal bounded region (UBR) by block size is extended in this paper by stating how they may be ordered.…”