Wentao Zheng

Robust tracking of soccer players based on data fusion

Presents a technique for integrating multiple visual features for tracking moving objects. Our proposed method consists of observation (pattern-matching) units and prediction units, which form a ladder structure. The major feature of our proposed method is that each of the observation units with different pattern matching algorithms is executed step-by-step to innovate the state vector considering the reliability of the observation. The fusion of multiple observations makes the tracks robust to occlusion and to deformation. Experiments with soccer sequences are shown to validate the techniques robustness. Its applications to broadcasting services are also briefly discussed.

Analysis of space-dependent characteristics of motion-compensated frame differences based on a statistical motion distribution model

Although it has been observed that motion-compensated frame differences increase toward block boundaries and overlapped block motion compensation (OBMC) has been shown to provide reduced blocking artifacts as well as improved prediction accuracy, there is almost no satisfactory theoretical basis that clearly interprets the space-dependent characteristics of motion-compensated frame differences, nor have the theoretical aspects of OBMC been investigated thoroughly. We first interpret the space-dependent characteristics of motion-compensated frame differences based on a novel statistical motion distribution model. We then apply the statistical motion distribution model to the analysis of prediction efficiency of OBMC. Through the analysis, we prove theoretically that OBMC can reduce and equalize the motion-compensated frame differences across a block. The analytical results are justified by empirical experiments with typical image sequences.

A high-precision camera operation parameter measurement system and its application to image motion inferring

Information about camera operations such as zoom, focus, pan, tilt and dollying is significant not only for efficient video coding, but also for content-based video representation. In this paper we describe a high-precision camera operation parameter measurement system and apply it to image motion inferring. First, we outline the implemented system which is designed to provide camera operation parameters with a high precision required for image coding applications. Second, we calibrate the camera lens to determine its exact optical properties, A pin-hole camera model with the 2nd order radial lens distortion and a two-image calibration technique are employed. Finally, we use the pan, tilt and zoom parameters measured by the system to infer image motion. The experimental results show that the inferred motion coincides with the actual motion very closely. Compared to the motion analysis techniques that estimate camera motion from video sequences, our approach does not suffer from ambiguity, thus can provide reliable and accurate image global motion. The obtained motion can be applied to image mosaicing, moving object segmentation, object-based image coding, etc.

Robust depth-map estimation from image sequences with precise camera operation parameters

The depth-map of a scene conveys the fundamental information that is extremely useful for object segmentation, compact representation of videos, etc. We present an algorithm for robust depth-map estimation from image sequences with precise camera operation parameters. This algorithm exploits precise camera operation parameters to find correspondence points among multiple frames of an image sequence of a static scene. Our algorithm is able to deal with arbitrary camera movement. The experimental results show significant improvement in the accuracy of depth-maps compared with the conventional two-frame matching method.

Analysis of space-dependent characteristics of motion-compensated frame differences

To design an optimal coding algorithm, a signal-source model that accurately reflects the signal characteristics is required. Although many investigations have been made on various aspects of motion-compensated frame differences, most of them assume that motion-compensated frame differences are homogeneous. Recently, however, it was observed that the variances of motion-compensated frame differences at block boundaries tend to be larger than those at block centers. In this paper we present a theoretical interpretation of the space-dependent characteristics of motion-compensated frame differences. The analysis is based on a statistical model for motion. The analytical results are justified by empirical experiments with typical image sequences.