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

Abstract

In the context of motion estimation for video coding, combining a successive elimination algorithm (SEA) with a motion estimation algorithm reduces the number of computed cost functions without impact on rate or distortion. The SEAs use the sum of absolute differences to eliminate motion vector candidates in the search area that cannot improve the current minimum. The novelty in this paper is that instead of relying on a static geometric pattern (i.e., like a spiral), we proposed a dynamic algorithm that creates a cost-based search orderings. A dynamic costbased search ordering not only improves elimination but also allows for early termination which removes, on average, 61% of the block-matching loop iterations performed by the rate-constrained successive elimination algorithm (RCSEA). Our experiments show that the proposed solution is 5 times faster than the high efficiency video coding (HEVC) HM encoder software in full search mode with a 0.02% impact on BD-Rate. This is twice the speed of the HEVC HM software encoder using only the RCSEA.

Multi-level rate-constrained successive elimination algorithm tailored to suboptimal motion estimation in HEVC

“…This follows the assumption that the likelihood of finding the best-matched candidate would decrease with an increase in the cost of the MV [27]. However, as explained in [28], these small refinement search areas are prone to be off-centered. When a search ordering based on a static geometric pattern is used on an off-centered search area, it will not follow the previous assumption, as the center of the refinement zone will probably not be the predicted MV.…”

“…This situation is ideal for the cost-based search ordering approach we proposed in [28]. That search ordering is dynamic and adapts to off-centered search areas, in addition to guaranteeing that candidates will be evaluated by increasing MV costs.…”

“…Even if the refinement search area is small (9 × 9), the MV cost remains significant because the predicted MV is not always the center of the search area; while the area itself is small, that might not be the case for the MV costs. Moreover, as explained in [28], the cost-based ordering maximizes the impact of the rate constraint on transitive elimination.…”

Relying on a Rate Constraint to Reduce Motion Estimation Complexity

“…Unfortunately, this method cannot be generalized for different algorithms because it operates in a specific TZS step. Meanwhile, the work in [12] leveraged the assumption that the likelihood of finding optimal candidates decreases as the MV bitrate increases to produce a rate-ordered variant of Successive Elimination [13]. Their algorithm provides the optimal solution (in terms of coding efficiency) within the search set, but limits the complexity reduction that can be achieved.…”

“…In light of the observations brought up by [11] and [12], that selected MV distribution and MV bitrate surfaces should be taken into account when designing a ME search pattern, we explore the impact of the MV bitrate estimation on fast IME algorithms. Thus, our contributions are: 3.…”