|
Résumé :
|
The digital video coding technology is growing more and more. In addition to the known applications needing more refinement, there are new applications (for example: Emails, mobile telephones of third generation, teleconferences…) which push the video coding research to develop more. All video data need to be compressed. Images are the principal element in video and their compression allows very important applications. However, the compression of each frame (image) separately is not optimal; the redundancy between successive frames must be considered. Motion estimation is a technique based on the compression of frames by referencing others. It is used in the famous standards such as MPEG, H26x. Motion estimation searches current frame objects in the reference one (past or future). The most used searching methods are the Block Matching algorithms owing to their performance and adaptability for hard implementations. A block matching algorithm subdivides the frame into non-overlapped rectangular blocks and searches these blocks in the reference frame. In this work, a motion estimation based video codec is designed and implemented for the application, study and comparison of motion estimation algorithms. A hybrid search technique is proposed too, based on block motion type detection from the information gathered from the precedent estimations. The block motion type (stationary, slow or large) is used to choose the appropriate algorithm to apply in search. The hybrid technique is based also on search starting point estimation instead of considering a large searching zone
|
