Summary: | 碩士 === 國立嘉義大學 === 資訊工程學系研究所 === 95 === Comparing to the previous coding standard, H.264 provides many schemes for improving the encoding result. One of the improving scheme is mode decision. Mode decision provides different block sizes (16x16,16x8,8x16,8x8,8x4,4x8, 4x4) to suit different video property. By the way, H.264 standard can save coding bits and gains a better video quality at the same time. On the contrary, the mode decision scheme makes coding complexity increase by times. The goal of this study is to maintain the compression quality and efficiently reduce the mode decision complexity.
In our mechanism, the effectiveness is focused on the concise utilization to the coding modes of two neighbors of an encoding MB to predict its coding mode with few candidates. This depends on two cognitions in the observation of splitting seams partitioning MB for ME. The first cognition will realize the splitting seam in the prediction MB as the type of boundary of needing consecutive stretch once again between two distinct textures. This cognition holds only under the condition that the direction of splitting seam and the direction of prediction MB’s location relative to encoding MB are same. Thus, according to the nature of boundary extending, the mechanism can choose the suited candidates from all coding modes available in H.264. On the contrast, if the direction of splitting seam and that of prediction MB’s location relative to encoding MB are perpendicular, the splitting seam in prediction MB is conceived with the second cognition. From the aspect of second cognition, the splitting seam behaves like a watershed to identify two regions of displaying different textures.
In this study, we use two cognitions to predict candidate coding modes. We success reduce the complexity of coding mode decision mechanism with small influence of compression quality.
|