An adaptive search ordering for rate-constrained successive elimination algorithms

Abstract

This paper proposes a solution for the problem of unnecessary cost function evaluations, found when combining the successive elimination algorithm with a spiral scan search ordering. Our experiments show that the implementation of such a combination inside the HEVC reference software leads to unnecessary cost function evaluations. On the tested video sequences, an average of 3.46% unnecessary cost function evaluations was measured. Considering only small block sizes (e.g., 4×8 and 8×4), this average rises to 8.06%. To solve this problem, we propose an adaptive scan ordering of block matching candidates within the search area. When used with our early termination threshold, the proposed approach will only evaluate necessary cost functions, without impacting ratedistortion.

Cost-Based Search Ordering for Rate-Constrained Motion Estimation Applied to HEVC

“…Concretely, geometric patterns are obsolete; the next candidate must be determined by its properties. In this work, the next candidate is determined by its cost, but it can be determined by other properties, such as its ADS [19].…”

“…Implementations of the RCSEA have been presented for H.264/MPEG-4 AVC in [17] and [18]. To our knowledge, except for our previous work [19], [20], no work has been published on RCSEA for H.265/HEVC. Althought the work of Seidel et al [21] also targets HEVC, it targets sub pel motion estimation whereas this work targets full pel motion estimation.…”

Sub-partition reuse for fast optimal motion estimation in HEVC successive elimination algorithms

“…We compare the proposed solution to our previous work [10] adapted to HEVC, and implemented in version 16.6 of the HM reference encoder. We did not compare it to [11], as speedups would be biased, because of the time required to sort candidates by ascending ADS.…”

“…Implementations of this algorithm have been presented for H.264 [8], [9]. To our knowledge, except for our previous works presented in [10] and [11], no work has been published so far on RCSEA for HEVC.…”

“…In [11], we showed that the smallest codomain is obtained by sorting the candidates by ADS value. In this work, we reduce the codomain of square prediction units (PUs) even more, by reusing information from previous ME operations performed over rectangular PUs.…”