| Summary: | 碩士 === 國立成功大學 === 工程科學系碩博士班 === 95 === In this paper, we propose the method to combine the use of motion vector embedded in MPEG bitstream to generate ”motion flow”, and the use of camera motion parameter of camera motion estimation to generate “camera motion sequence”, which are applied them to perform quick video retrieved. We want to implement the system which we can retrieved the video in compressed domain without segmenting the moving object..
There are three sections in our experiment. One is retrieved by “camera motion sequence”, One is retrieved by “motion flow”, and the last is retrieved by both these features. From the experiment, we can know that the retrieval by motion flow is better. The precision of our method is 50%, but the recall is 100%, and we can use the system to do more detail searching.
|
|---|
