Deblocking Filter Thresholds Using Sub-Pel Motion Vectors

The patent changes the motion vector threshold parameter from full-pel accuracy (integer pixels) to sub-pel accuracy (fractional pixels, e.g., 1/4 pel or 1/8 pel). This parameter change allows the deblocking decision to be more sensitive to motion differences, enabling effective blocking artifact reduction even when using lower motion accuracy prediction modes, while maintaining computational efficiency through the use of fixed-point arithmetic representations.

The patent replaces the traditional floating-point or complex fixed-point arithmetic operations with a simplified fixed-point arithmetic system. By representing sub-pel motion vectors and thresholds using integer fixed-point formats (e.g., Q4.12 or Q8.16), the patent eliminates the need for complex floating-point units while achieving sub-pel precision, thus reducing hardware complexity and computational overhead.

The patent changes the precision parameter of motion vector threshold from full-pel to sub-pel accuracy, enabling more precise measurement of motion differences between adjacent blocks. This allows the deblocking filter to make more accurate decisions about where blocking artifacts are likely to occur, improving measurement precision while the fixed-point implementation keeps computational overhead manageable.