Chaur-Heh Hsieh

VLSI architecture for block-matching motion estimation algorithm

The block-matching motion estimation is the most popular method for motion-compensated coding of image sequence. A VLSI architecture for implementing a full-search block-matching algorithm is presented. Based on a systolic array processor and shift register arrays with programmable length, the proposed architecture has the following advantages: it allows serial data input to save the pin counts but performs parallel processing; it is flexible in adaptation to the dimensional change of the search area via simple control; it can operate in real time for videoconference applications; and it is simple and modular in design, and thus is suitable for VLSI implementation. >

Lossless compression of VQ index with search-order coding

In memoryless vector quantization (VQ) for images, each block is quantized independently and its corresponding index is sent to the decoder. This paper presents a new lossless algorithm that exploits the interblock correlation in the index domain. We compare the current index with previous indices in a predefined search path, and then send the corresponding search order to the decoder. The new algorithm achieves significant reduction of bit rates without introducing extra coding distortion when compared to memoryless VQ. It is very simple and computationally efficient.

Fast search algorithms for vector quantization of images using multiple triangle inequalities and wavelet transform

The encoding of vector quantization (VQ) needs expensive computation for searching the closest codevector to the input vector. This paper presents several fast encoding algorithms based on multiple triangle inequalities and wavelet transform to overcome this problem. The multiple triangle inequalities confine a search range using the intersection of search areas generated from several control vectors. A systematic way for designing the control vectors is also presented. The wavelet transform combined with the partial distance elimination is used to reduce the computational complexity of the distance calculation of vectors. The proposed algorithms provide the same coding quality as the full search method. The experimental results indicate that the new algorithms perform more efficiently than existing algorithms.

Design of two-dimensional FIR digital filters by a two-dimensional WLS technique

For two-dimensional (2-D) FIR filter design, the conventional weighted least squares (WLS) technique rearranges the filter parameters of 2-D form into their corresponding one-dimensional (1-D) form, thus resulting in expensive computation. This paper presents a new computationally efficient WLS technique for the design of 2-D FIR filters. We introduce an updating desired frequency response which implicitly includes the weighting function such that the sum of weighted square errors to be minimized can be represented in a 2-D matrix form. This makes it possible to keep all filter parameters in their natural 2-D form, thereby reducing the computational complexity from O(N/sup G/) to O(N/sup 3/). It is confirmed through design examples that the new technique is computationally very efficient and leads to nearly optimal approximations. This technique is suitable for the design of 2-D real zero-phase FIR filters with quadrantal symmetric or antisymmetric frequency response and can also be applied to the design of 1-D FIR filters.

Motion estimation for video compression using Kalman filtering

Motion estimation plays an important role for the compression of video signals. This paper presents a new block-based motion estimation method using Kalman filtering. The new method utilizes the predicted motion and measured motion to obtain an optimal estimate of motion vector. The autoregressive models are employed to fit the motion correlation between neighboring blocks and then achieve predicted motion information. The measured motion information is obtained by the conventional block-based fast search schemes. Several algorithms based on either one- or two dimensional models using either nonadaptive or adaptive Kalman filters are developed. The analysis of computational complexity and the simulation results indicate that the proposed method achieves significant savings on computation along with smoother motion vector fields and similar picture quality, when compared to the conventional full search algorithm.

A Novel Prediction-Based Directional Asymmetric Search Algorithm for Fast Block-Matching Motion Estimation

This paper proposes a very fast block-matching motion estimation algorithm for video compression. This method uses a new concept involving a very compact center-biased characteristic in developing directional asymmetric search patterns, which we refer to as directional asymmetric search (DAS). The initial pattern of the DAS is a compact cross pattern containing only five initial search points. The DAS utilizes error information (block distortions) of the search patterns to determine the search direction, and then asymmetric search patterns are used in the subsequent steps accordingly. Furthermore, a prediction scheme and a best match prejudgment scheme are incorporated to favor fast motion and to benefit stationary and quasi-stationary blocks, respectively. Therefore, the proposed method significantly reduces the number of search points for locating a motion vector. Compared to conventional fast algorithms, the proposed method has the fastest search speed and most satisfactory PSNR values for all test sequences.

Real-time image template matching based on systolic array processor

An architecture based on a systolic array for real-time image template matching is presented. The architecture consists mainly of four elements: a digitizer, a two-dimensional systolic array combined with variable-length shift register arrays, an adder tree, and a comparator. All the elements form a four-stage pipeline. The image data enter the pipe sequentially in the same order as the TV raster scan. The matching computation is, however, performed in a parallel manner. The analyses on time complexity and hardware complexity have shown that real-time performance is achieved. The analyses have also shown that the processing speed is higher and the hardware is simpler when compared to the architecture presented by Chou and Chen.

Fast codebook generation algorithm for vector quantization of images

Abstract A partial search partial distortion (PSPD) algorithm is presented for reducing the computational complexity of the codebook generation in vector quantization (VQ) of images. A partial codebook is built up using the feature of a training vector. The partial distortion technique is employed to find the best match codevector from the partial codebook. The new algorithm results in a considerable reduction of computational complexity when compared to the conventional LBG algorithm.

Event Detection of Broadcast Baseball Videos

This paper presents an effective and efficient event detection system for broadcast baseball videos. It integrates midlevel cues including scoreboard information and shot transition patterns into event classification rules. First, a simple scoreboard detection and recognition scheme is developed to extract the game status from videos. Then, a shot transition classifier is designed to obtain the shot transition patterns, which contains several novel schemes including adaptive playfield segmentation, pitch shot detection, field shot detection, as well as infield/outfield classification. The extracted midlevel cues are used to develop an event classifier based on a Bayesian Belief Network. The network is with low complexity because the number of these cues used is small, which not only improves the generalization performance of the event classifier but also reduces system complexity as well as training efforts. Using the inference results of the network, we further derive a set of classification rules to identify baseball events. The set of rules is stored in a look-up table such that the classification is only a simple table look-up operation. The proposed approach is very simple and computational efficient. More importantly, the simulation results indicate that it identifies ten significant baseball events with 95% of precision rate and 92% of recall rate, which is very promising.

Block-matching motion estimation using correlation search algorithm

Block-matching motion estimation plays an important role in video coding. In general, there exists a high motion correlation between neighbor blocks in spatial and temporal directions. This paper presents a novel motion estimate scheme, called correlation search, which attempts to find the highest motion-correlation neighbor block from the spatial and temporal neighbor blocks. The motion vector of the highest motion-correlation neighbor block is regarded as the motion estimate of the current block. The correlation search scheme can be based on any existing block matching algorithms. When the scheme is based on a full-search algorithm, it achieves almost the same estimate accuracy with a significant reduction of computational complexity. If the scheme is based on a fast search algorithm, it obtains better estimate accuracy in addition to the improvements in computation.