Adaptive Motion Vector Precision for Video Coding

The patent applies different motion vector precision levels to different video content types. Specifically, it uses higher precision (e.g., 1/4 pixel or 1/8 pixel) for content that benefits from it (such as natural video with smooth motion) and lower precision (e.g., integer pixel) for content where high precision provides minimal benefit (such as screen content with sharp edges). This local differentiation resolves the contradiction by optimizing precision per content type rather than applying a uniform standard.

The patent dynamically changes the motion vector precision parameter based on content type detection. The system detects whether the video content is natural video, screen content, or mixed content, and adjusts the MVD precision accordingly. This parameter adaptation allows the system to achieve high precision where needed while maintaining low precision where sufficient, thereby optimizing the bitrate-accuracy tradeoff.

The patent implements a dynamic selection mechanism that chooses among multiple MVD coding techniques (such as CABAC, CAVLC, or other entropy coding methods) based on the specific content being encoded. Rather than implementing all techniques simultaneously or using a fixed technique, the system dynamically selects the most appropriate coding method for each block or sequence based on content characteristics, resolution, and other factors. This dynamic approach improves coding efficiency while managing complexity through selective implementation.

The patent divides the video content into different segments or blocks that can be encoded using different MVD coding techniques. By segmenting the video stream and applying different coding methods to different segments based on their specific characteristics, the system achieves improved overall coding efficiency without requiring the entire system to handle all possible coding techniques simultaneously, thus managing device complexity.