Research On Fast Picture Adaptive Frame/Field Algorithom In Video Compression

Interlaced scanning technique has come into broad use since the born of television technique. Many precious video materials are scanned in interlaced-format; hence, it will be one of the main video formats for a long time. When picture adaptive frame/field (PAFF) is utilized to interlaced sequences, two fields can be combined into one frame, which is encoded in frame coding; besides, they can also be encoded in field coding separately. Adopting field or frame coding can be determined by coding performance adaptively. Although PAFF can enhance the coding performance, whereas, the time complexity ascends more than twice. In order to reduce time consumption, one-pass fast PAFF based on motion diction is proposed in this paper.This paper has carried on a deep study on PAFF in MPEG-2 and H.264. We have discovered that the switch between frame and field has a close relationship with the motion activity of images. Besides, PAFF in different standards has its own characteristics. The two points mentioned above can be used as guiders for the design of the algorithm in this paper.The sum of square differences (SSD), which has been defined as square summation of corresponding pixel differences between current and previous frame, has been endowed a novel connotation in this paper. The selection of frame and field has been tested in standard-definition sequences. The experimental results indicated that selection of frame and field in AVS has four characteristics, which are continuity of frame/field selection, consistency of SSD curves, relativity of motion information, as well as pertinence of quantization steps. According to the aforementioned characteristics, the core algorithms proposed in this paper is that selection of frame/field is determined before encoding, where field coding is chosen when motion is acute, whereas, frame coding is chosen when motion is mild, and the coding scenario of the previous frame is applied for other circumstance. The consistency of the SSD curve indicates that the change of SSD can describe the acute extent of motion, however, due to the relax distribution; it is hard to use directly. In this paper, SSD is normalized, which is to say, the change rate of the SSD of the current...