| Summary: | 博士 === 國立臺灣大學 === 電機工程學研究所 === 89 === Abstract
This dissertation presents three video processing techniques that are essential for real-time MPEG-4 encoding systems.
The content-based interactivity is one of the most important features of the MPEG-4 visual coding; however, the automatic generation of the shape information is still an open research problem. In the first part of this dissertation, an efficient moving object segmentation is proposed to automatically generate the shape information. Background registration technique is used in this method to construct a reliable background image from the accumulated frame difference information. The moving object region is then separated from the background region by comparing the current frame with the constructed background image. Finally, a post-processing step is applied on the obtained object mask to remove noise regions and to smooth the object boundary. In situations where object shadows appear in the background region, a pre-processing gradient filter is applied on the input image to reduce the shadow effect. Very good segmentation quality and high execution speed are demonstrated by the experimental results. This algorithm is very suitable for real-time MPEG-4 VOP generation.
The motion estimation is always the most computationally intensive part of a video encoding system. In the second part of the dissertation, an efficient motion estimation algorithm for real-time implementations of MPEG-4 encoder on multimedia processors is presented. The motion vector predictor position is used as the starting point in the search process because the correlation between neighboring motion vectors is strong. The line search pattern is used in the proposed algorithm to reduce the memory access as well as to exploit the special multimedia processor instructions for SAD calculations. Experimental results show that the performance of the proposed predictive line search (PLS) is very close to the full search algorithm. Compared with the well-known diamond search fast algorithm, the predictive line search shows better performance and robustness especially for high motion sequences.
In order to exploit the concept of system-on-a-chip, a CMOS active pixel sensor (APS) camera chip with direct frame difference output is reported in the last part of the dissertation. This chip combines an important signal processing for video segmentation with the image sensing circuit to form a smart camera system. The proposed APS cell circuit has in-pixel storage for previous frame image data so that the current frame image and the previous frame image can be read out simultaneously in differential mode. The signal swing of the pixel circuit is maximized for low supply voltage operation. Good image quality and correct signal processing results can be obtained from the experimental results.
These techniques can greatly help the design and implementation of real-time MPEG-4 applications.
|
|---|
