9:30, IMDSP-P3.1
A MATCH MOVING TECHNIQUE FOR HUMAN VIDEO SEQUENCES
J. HOSHINO, H. SAITO
Merging virtual object with human video sequence is an important technique for many applications such as special effects in movies and augmented reality. In a traditional method, the operator manually fit 3D body model onto the human video sequence, and generate virtual objects at the current 3D body pose. However, the manual fitting is a time consuming task, and the automatic registration is required. In this paper, we propose a new method for merging virtual objects onto the human video sequence. First, we track the current 3D pose of human figure by using the spatio-temporal analysis and the structural knowledge of human body. Then we generate CG objects and merge it with the human figure in video. In this paper, we demonstrate examples of merging virtual cloth with the video captured images.

9:30, IMDSP-P3.2
A NOVEL HEXAGON-BASED SEARCH ALGORITHM FOR FAST BLOCK MOTION ESTIMATION
C. ZHU, X. LIN, L. CHAU, K. LIM, H. ANG, C. ONG
In block motion estimation, search pattern with different shape or size has very important impact on search speed and distortion performance. In this paper, we propose a novel algorithm using hexagon-based search (HEXBS) pattern for fast block motion estimation. The proposed HEXBS algorithm may find any motion vector with fewer search points than the diamond search (DS) algorithm. The speedup gain of the HEXBS method over the DS algorithm is more striking for finding large motion vectors. Experimental results substantially justify the fastest performance of the HEXBS algorithm compared with several other popular fast algorithms.

9:30, IMDSP-P3.3
WAVELET VIDEO COMPRESSION USING REGION BASED MOTION ESTIMATION AND COMPENSATION
D. LAZAR, A. AVERBUCH
We present wavelet based video compression algorithms. The motion estimation and compensation, which is an essential part in the compression, is based on segment movements. The proposed based codec sequentially employs the following steps: We first divide each frame in the sequence into blocks resulted from the block matching algorithm. The, we appply Hough transform, in order to group it into segments the blocks that have similar motion parameters. We estimate the motion parameters for each segment using several motion models and least-squares algorithm and apply motion compensation upon the frames in each group. In the wavelet domain an optimal vector bit allocation is being used to distribute the bit budget among consecutive frames. The manipulated frames in each group are quantized and coded using techniques from still image compression. The residual error of the compression is reduced by applying error correction in the wavelet domain to enhance the quality of the reconstructed sequence.

9:30, IMDSP-P3.4
GENERALIZED PARTIAL DISTORTION SEARCH ALGORITHM FOR FAST BLOCK MOTION ESTIMAION
C. CHEUNG, L. PO
In this paper, a novel block-matching algorithm for fast motion estimation named generalized partial distortion search algorithm (GPDS) is proposed. It uses halfway-stop technique with progressive partial distortion (PPD) to increase the chance of early rejection of impossible candidate motion vectors at very early stages. Simulations on PPD show that 28 to 38 times computational reduction with only 0.45-0.50dB PSNR performance degradation compared to full search algorithm (FS). A new normalized partial distortion comparison method is also proposed for enabling the control of speed against quality by a speedup factor k. It generalizes the conventional partial distortion search algorithm when k is equal to 1, and the normalized partial distortion search algorithm (NPDS) when k is equal to infinity. Simulations show that GPDS with PPD provides PSNR performance very close to FS and NPDS with 7 to 17 times and 22 to 33 times speedup, respectively, as compared to FS.

9:30, IMDSP-P3.5
MEDIAN FILTERING-BASED PYRAMIDAL MOTION VECTOR ESTIMATION
J. ZAN, M. AHMAD, M. SWAMY
In this paper, a median filtering-based hierarchical motion vector estimation scheme making use of a pyramidal data structure is proposed. Compared to the conventional hierarchical motion vector estimation schemes, the proposed scheme overcomes the problem of propagation of false motion vectors across resolutions. Simulation studies show that the proposed scheme not only improves the prediction accuracy with respect to the prediction mean square error, but also results in a smoother motion field, which can be encoded with less number of coding bits. It is shown that an improvement in the rate distortion performance is achieved with little increase in the computational complexity. It is also shown that Burt and Adelson's pyramidal data structure provides the best performance among a number of the generating kernels considered in our study.

9:30, IMDSP-P3.6
LOCAL BANDWIDTH CONSTRAINED FAST INVERSE MOTION COMPENSATION FOR DCT-DOMAIN VIDEO TRANSCODING
S. LIU, A. BOVIK
DCT-based digital video coding standards such as MPEG and H.26x are becoming more widely adopted for multimedia applications. Since the standards differ in their format and syntax, video transcoding, where a compressed video bitstream is converted from one format to another format, is of interest for purposes such as channel bandwidth adaptation and video composition. DCT-domain video transcoding is generally more efficient than spatial domain transcoding. However, since the data is organized block by block in the DCT-domain, inverse motion compensation becomes the bottleneck for DCT-domain methods. In this paper, we propose a novel local bandwidth constrained fast inverse motion compensation algorithm operating in the DCT-domain. Relative to Chang’s algorithm [1], the proposed algorithm achieves computational improvement of 25% to 55% without visual degradation. A by-product of the proposed algorithm is a reduction of blocking artifacts in very low bit-rate compressed video sequences.

9:30, IMDSP-P3.7
A MULTI-STAGE MULTI-CANDIDATE ALGORITHM FOR MOTION ESTIMATION
T. LIAO, S. PHOONG, Y. LIN
Motion compensation using the optimal full search algorithm is often too computational heavy for real time implementation. Many suboptimal fast search algorithms have been proposed. In particular, Liu and Zaccarin proposed the Alternating Subsampling Search Algorithm (ASSA). The ASSA reduces the computation by subsampling the pixels instead of limiting the search locations. It was shown that ASSA has nearly the same MSE performance as the full search but its complexity is only 1/4 of the full search. In this paper, we generalize the idea to the multi-stage case. Simulation results show that the proposed algorithm has a comparable performance to the ASSA but it has a much lower computational cost.

9:30, IMDSP-P3.8
BEST VIEW SELECTION AND COMPRESSION OF MOVING OBJECTS IN IR SEQUENCES
N. VASWANI, R. CHELLAPPA
A system for selection of a single best view image chip from an IR video sequence and compression of the chip for transmission is reported. Moving object detection was done using the algorithm described in [1]. Eigenspace classification has been implemented for best view selection. Fast algorithms for image chip compression have been developed in the wavelet domain by combining a non-iterative zerotree coding method with 2D-DPCM for both low and high frequency subbands and compared against existing schemes.

9:30, IMDSP-P3.9
2-D MOTION ESTIMATION WITH HIERARCHICAL CONTENT-BASED MESHES
G. AL-REGIB, Y. ALTUNBASAK
Two-dimensional mesh-based models provide a good alternative to motion estimation and compensation. The estimation of the best motion vectors at the node-point motion vectors is a challenging task. To this effect, Nakaya et al. proposed a hexagonal matching procedure. Toklu et al. improved the hexagonal search algorithm in terms of both motion estimation accuracy and computational complexity by employing a hierarchy of regular meshes. Recognizing the limitations of regular meshes, Van Beek et al. extended Toklu's work by employing content-based meshes. Here, we provide an alternative hierarchical motion estimation method with content-based meshes where hierarchical representations are employed for the images as well as meshes in order to provide further improvements in computational complexity as well as motion accuracy. The comparison results are provided with real video sequences.