H.264/AVC 비트스트림 영역에서의 부분복원을 이용한 다중객체의 실시간 검출 및 추적

소개

In this paper, we show that we can apply probabilistic spatiotemporal macroblock filtering (PSMF) and partial decoding processes to effectively detect and track multiple objects in real time in H.264|AVC bitstreams with stationary background. Our contribution is that our method cannot only show fast processing time but also handle multiple moving objects that are articulated, changing in size or internally have monotonous color, even though they contain a chaotic set of non-homogeneous motion vectors inside. In addition, our partial decoding process for H.264|AVC bitstreams enables to improve the accuracy of object trajectories and overcome long occlusion by using extracted color information.

목차

Abstract

1. Introduction

2. Related Works
2.1. Clustering-based Approaches
2.2. Filtering-based Approaches

3. Probabilistic Spatiotemporal Macroblock Filtering
3.1. Block Clustering
3.2. Spatial Filtering
3.3. Temporal Filtering

4. Refinement and Recovery of Object Trajectories
4.1. Partial Decoding and Background Subtraction
4.2. Motion Interpolation
4.3. Occlusion Handling

5. Experimental Results
5.1. Video Sequences
5.2. Extraction Phase
5.3. Refinement Phase
5.4. Performance Analysis and Occlusion Handling
5.5. Computational Cost

본문내용

In this paper, we have presented a novel approach that can rapidly detect and track multiple objects simultaneously from H.264|AVC bitstreams. Our proposed algorithm is considerably practical since it does not merely work well in real time even in general PCs, but also it can efficiently adapt to more natural scenes which include multiple objects, long-time occlusion or the movement of any articulated object. The flexibility is a novelty of the algorithm in that no previous compressed domain algorithms have been verified in such complicated environments.
These features are accomplished by combining the compressed domain approach, which roughly extracts the macroblock-level object region based on probabilistic spatiotemporal macroblock filtering, with the pixel domain approach which precisely refines the object trajectories based on partially decoded color information. In particular, the proposed partial decoding scheme for H.264|AVC bitstreams is proven to be significantly effective in reducing the computational cost. Therefore, this kind of combination can be employed as a fundamental framework to solve the problem of slow processing from which existing vision-based algorithms of pixel domain have suffered. We expect that the proposed algorithm is applicable to many real-time surveillance systems.

참고자료

· 없음