Parameterized Video Resampling Filters for Low-Bitrate Quality

The patent applies dynamics by making the filter selection adaptive rather than fixed. The encoder dynamically selects from multiple filter types (e.g., bilinear, bicubic, Lanczos) based on local image characteristics such as edge orientation, texture complexity, and frequency content. This allows the compression system to optimize visual quality at each region while maintaining overall low bit rate, resolving the contradiction between reduced bit rate and maintained visual quality.

The patent implements local quality by applying different filtering strategies to different regions of the video frame. High-frequency regions with edges or fine details receive more sophisticated filtering (e.g., bicubic or Lanczos) to preserve quality, while low-frequency smooth regions use simpler filtering (e.g., bilinear) to save bits. This regional adaptation maintains visual quality where needed while achieving overall bit rate reduction.

The patent applies parameter changes by varying filter order, kernel size, and cutoff frequencies based on local image characteristics. In regions requiring high-frequency preservation, the system uses higher-order filters with larger kernels and less aggressive low-pass filtering. In regions where quality is less critical, it uses lower-order filters with smaller kernels and stronger compression. This dynamic parameter adjustment maintains quality where needed while controlling overall bit rate.

The patent implements partial action by applying high-quality filtering only to necessary regions rather than uniformly across the entire frame. By identifying regions with edges, textures, or fine details that require preservation, the system applies sophisticated filtering selectively to those areas while using aggressive compression in smooth, less critical regions. This partial application of high-quality processing maintains essential visual information while keeping the overall bit rate low.

The patent applies segmentation by dividing the video frame into multiple regions based on local characteristics such as edge detection, gradient magnitude, and frequency analysis. Each segment is then assigned an appropriate filter type from the available set (bilinear, bicubic, Lanczos, etc.). This segmentation approach allows the system to use complex filtering only where necessary while using simpler filters elsewhere, thereby improving visual quality without uniformly increasing computational complexity across the entire frame.

The patent implements dynamics by making filter selection adaptive based on real-time analysis of local image characteristics. Rather than using a fixed filter for the entire frame or pre-defining static regions, the system dynamically determines the appropriate filter type for each region based on current image content. This dynamic adaptation improves visual quality where needed while avoiding unnecessary computational complexity in regions where simple filtering suffices.