Changryoul Choi

New sorting-based partial distortion elimination algorithm for fast optimal motion estimation

In this paper, we present a new sorting-based partial distortion elimination (PDE) algorithm for fast optimal motion estimation. By analyzing the contributions to the true sum of absolute differences (SAD), we found that there is a close relationship between the distances from the mean value of the current block and the contributions to the true SAD. By sorting the distances from the pixels to the current block and subtracting them from a mean value, then applying this order to the typical PDE, we can eliminate impossible candidates faster and save substantial computations. Experimental results show that the proposed algorithm reduces computational complexity by about 45% on average compared with the typical PDE.

Enhanced two-bit transform based motion estimation via extension of matching criterion

Improved two-bit transform-based motion estimation algorithms are proposed in this paper. By extending the typical two-bit transform (2BT) matching criterion, the proposed algorithms enhance the motion estimation accuracy with almost the same computational complexity, while preserving the binary matching characteristic. Experimental results show that the proposed algorithm achieves peak-to-peak signal-to-noise ratio (PSNR) gains up to 0.42 dB compared with the conventional 2BT-based motion estimation.

Successive elimination algorithm for two-bit transform-based motion estimation

A successive elimination algorithm for two-bit transform (2BT) based motion estimation (ME) is proposed. By mathematically deriving the lower bound for2BT-based matching criterion, we can discard the impossible candidates earlier and save computations substantially. Experimental results show that although the performance of the proposed algorithm is the same as that of the full search 2BT (FS-2BT) based ME algorithm, the computational complexity has been reduced significantly.

Modified rate distortion optimization using inter-block dependence for H.264/AVC intra coding

H.264/AVC allows several prediction schemes including variable modes for intra and inter predictions. To take full advantage of all allowed modes, H.264 employs rate distortion optimization (RDO) as its mode decision method. However, RDO incorporated into the H.264 encoder assumes block independence because calculating rate distortion (RD) cost for all possible mode sets requires a huge number of computations. In contrast to this assumption, coding efficiency of intra prediction in the H.264 encoder is affected by reference pixels in neighboring blocks. Thus, RDO in the H.264 encoder has potential to improve coding performance by considering the accuracy of intra prediction for the next blocks during the mode decision process of the current block. In this paper, we propose new criteria for I4 times 4 mode decision and by using this criteria, the proposed method considers more mode sets to improve RD performance compared to the original RDO. Although the proposed method requires more computations compared to the original RDO, experimental results indicate that the proposed method noticeably improves coding efficiency compared to the original RDO.

Constrained two-bit transform for low complexity motion estimation

In transforming the original image frames into two-bit representations, the typical two-bit transform (2BT) needs to calculate the variances of the local blocks. This calculation of variances inevitably involves multiplication operations and renders the computational complexity of typical 2BT somewhat high. In this paper, we propose a constrained two-bit transform (C2BT) for low complexity motion estimation (ME). By exploiting the advantages of the typical constrained one-bit transform (C1BT) and the typical 2BT, the proposed algorithm significantly reduces the computational complexity of transformation of image frames into two-bit representations. Experimental results show that the proposed algorithm enhances the ME accuracy by 0.35dB and 0.28dB compared with the 2BT-based ME and the C1BT-based ME, respectively.

Improved two-bit transform-based motion estimation via extension of matching criterion

An improved two-bit transform-based motion estimation algorithm is proposed in this paper. By extending the typical two-bit transform (2BT) matching criterion, the proposed algorithm enhances the motion estimation accuracy with almost the same computational complexity, while preserving the binary matching characteristic. Experimental results show that the proposed algorithm achieves peak-to-peak signal-to-noise ratio (PSNR) gains of 0.29dB on average compared with the conventional 2BT-based motion estimation.

Successive Elimination Algorithm for Constrained One-bit Transform Based Motion Estimation Using the Bonferroni Inequality

The constrained one-bit transform (C1BT) was proposed to increase the motion estimation (ME) accuracy of the previous one-bit transform (1BT) especially for small motion blocks. Although making another bit-plane is very simple and efficient, its performance is even better than that of the two-bit transform (2BT) based ME. However, unlike the 1BT-based ME and the 2BT-based ME, the successive elimination algorithm (SEA) based on the triangle inequality for C1BT-based ME cannot be derived because C1BT matching error criterion does not satisfy the typical measure conditions. In this letter, a fast full-search block matching algorithm for C1BT-based ME is developed. The proposed algorithm evaluates lower bounds for constrained one-bit matching criterion based on the Bonferroni inequality to eliminate the impossible candidates faster and save computations substantially. Experimental results show that while the ME accuracy of the proposed algorithm is the same as that of the full search C1BT, the proposed algorithm reduces computational complexity significantly.

Low complexity weighted two-bit transforms based multiple candidate motion estimation

A low complexity weighted two-bit transforms (2BT) based multiple candidate motion estimation algorithm is proposed in this paper. By exploiting almost the identical operations in two different matching error criteria, we can efficiently determine two best motion vectors according to the respective matching criteria with negligible complexity increase. And with additional local square search, the proposed algorithm can enhance the overall motion estimation accuracy substantially. Experimental results show that the proposed algorithm achieves peak-to-peak signal-tonoise ratio (PSNR) gains about 0.55dB on average compared with the conventional 2BT-based motion estimation1.

Bit-inverted Gray coded bit-plane matching for low complexity motion estimation

In this paper, a bit-inverted Gray coded bit-plane matching algorithm is proposed for low complexity motion estimation. Unlike the typical Gray coded bit-plane matching algorithms, the proposed algorithm uses bit-inverted Gray codes for transforming image frames and a corresponding extended matching criterion to enhance the motion estimation accuracy. Experimental results show that the proposed algorithm outperforms other bit-plane matching based motion estimation algorithms while preserving the binary matching characteristic.

Fast motion estimation algorithm using dual bit-plane matching criteria

The constrained two-bit transform (C2BT) is recently proposed for low complexity motion estimation (ME) to reduce the high computational complexity of the typical two-bit transform (2BT) while maintaining the ME accuracy. And the bit-inverted Gray-coded bit-plane matching (BGCBPM) is also proposed to enhance the ME accuracy. In this paper, we propose a fast ME algorithm exploiting these two bit-plane matching (BPM) criteria. Using the low complexity image transformations of typical images into bit-planes and their corresponding matching criteria and another hybrid matching criterion, we can efficiently determine multiple candidate motion vectors and increase the ME accuracy substantially. Experimental results show that the peak-to-peak signal-to-noise ratio (PSNR) difference between the proposed algorithm and the typical sum of absolute differences (SAD) based full search algorithm is only 0.03 dB on average with negligible computational complexity increase.