During the encoding process we separate the bitrate to optimize it for each resolution. For the highest quality, we use the maximum value for all resolutions (310 MB) combined. For the lower quality we start at 300 MB for worst case, and improve up to 500 MB if possible. The bitrates for 720p and 1080p are identical, as the key difference in the encoded video content is the ratios of interlacing, and the accompanying artifacts that can be seen at higher resolutions. At higher resolutions, more than a raw bitrate can be achieved by allowing the encoder to use the highest possible bitrate with temporal complexity reduction, and by limiting the bitrate for the remainder of the frame.
For the highest quality parameter settings, we used the Vivid preset, the "science" preset, and Netflix's speed value. The "science" preset set the frame rate as 25 fps, converted the interlaced video to progressive, limited color bitdepth to 5 bits, and reduced the denoise strength.
For the settings used to create the lowest quality parameter settings, we used the Extreme preset, with the same settings, above, except for the vignette (vignette). The 62.5% setting is low, but we set it as high as possible with no visible artifacts as the reduction in bitrate is minimal and even smoother motion. The 38% setting, however, yields visible and still noticeable artifacts, and we used it for the lowest quality.
One important limitation of H.264 encoding is that the quality is only as good as the source file. Newer, higher quality raw or captured videos are able to produce even higher quality H.264 video, but older, lower-quality videos have more of a risk of producing visible encoding artifacts at higher resolutions. Common downscaling may also cause visible artifacts due to the loss of finer details in the video. d2c66b5586