text/microsoft-resx 2.0 System.Resources.ResXResourceReader, System.Windows.Forms, Version=2.0.0.0, Culture=neutral, PublicKeyToken=b77a5c561934e089 System.Resources.ResXResourceWriter, System.Windows.Forms, Version=2.0.0.0, Culture=neutral, PublicKeyToken=b77a5c561934e089 17, 17 Psychovisual Rate Distortion Optimization sure is a mouthful, isn't it? Basically, it means x264 tries to retain detail, for better quality to the human eye, as opposed to trying to maximize quality the way a computer understands it, through signal-to-noise ratios that have trouble telling apart fine detail and noise. x264 has a variety of algorithms to decide when to use B-frames and how many to use. Fast mode takes roughly the same amount of time no matter how many B-frames you specify. However, while fast, its decisions are often suboptimal. Optimal mode gets slower as the maximum number of B-Frames increases, but makes much more accurate decisions, especially when used with B-pyramid. After the encoder has done its work, it has a bunch of data that needs to be compressed losslessly, similar to ZIP or RAR. H.264 provides two options for this: CAVLC and CABAC. CABAC decodes a lot slower but compresses significantly better (10-30%), especially at lower bitrates. If you're looking to minimize CPU requirements for video playback, disable this option. Baseline profile, as required for iPods and similar devices, requires CABAC to be disabled. x264 normally zeroes out nearly-empty data blocks to save bits to be better used for some other purpose in the video. However, this can sometimes have slight negative effects on retention of subtle grain and dither. Don't touch this unless you're having banding issues or other such cases where you are having trouble keeping fine noise. Trellis fine-tunes the rounding of transform coefficients to squeeze out 3-5% more compression at the cost of some speed. "Always" uses trellis not only during the main encoding process, but also during analysis, which improves compression even more, albeit at great speed cost. Trellis costs more speed at higher bitrates H.264 has a built-in deblocking filter that smooths out blocking artifacts after decoding each frame. This not only improves visual quality, but also helps compression significantly. The deblocking filter takes a lot of CPU power, so if you're looking to minimize CPU requirements for video playback, disable it. The deblocking filter has two adjustable parameters, \"strength\" and \"threshold\". The former controls how strong (or weak) the deblocker is, while the latter controls how many (or few) edges it applies to. Lower values mean less deblocking, higher values mean more deblocking. The default is 0 (normal strength) for both parameters. H.264 has a built-in deblocking filter that smooths out blocking artifacts after decoding each frame. This not only improves visual quality, but also helps compression significantly. The deblocking filter takes a lot of CPU power, so if you're looking to minimize CPU requirements for video playback, disable it. The deblocking filter has two adjustable parameters, \"strength\" and \"threshold\". The former controls how strong (or weak) the deblocker is, while the latter controls how many (or few) edges it applies to. Lower values mean less deblocking, higher values mean more deblocking. The default is 0 (normal strength) for both parameters. The 8x8 transform is the single most useful feature of x264 in terms of compression-per-speed. It improves compression by at least 5% at a very small speed cost and may provide an unusually high visual quality benefit compared to its compression gain. However, it requires High Profile, which many devices may not support. Mode decision picks from a variety of options to make its decision: this option chooses what options those are. Fewer partitions to check means faster encoding, at the cost of worse decisions, since the best option might have been one that was turned off. This setting controls both subpixel-precision motion estimation and mode decision methods. Subpixel motion estimation is used for refining motion estimates beyond mere pixel accuracy, improving compression. Mode decision is the method used to choose how to encode each block of the frame: a very important decision. SAD is the fastest method, followed by SATD, RD, RD refinement, and the slowest, QPRD. 6 or higher is strongly recommended: Psy-RD, a very powerful psy optimization that helps retain detail, requires RD. 10, the most powerful and slowest option, requires trellis=2. This is the distance x264 searches from its best guess at the motion of a block in order to try to find its actual motion. Doesn't apply to Diamond or Hexagon search options. The default is fine for most content, but extremely high motion video, especially at HD resolutions, may benefit from higher ranges, albeit at a high speed cost. Controls the motion estimation method. Motion estimation is how the encoder estimates how each block of pixels in a frame has moved. A better motion search method improves compression at the cost of speed. Diamond: performs an extremely fast and simple search using a diamond pattern. Hexagon: performs a somewhat more effective but slightly slower search using a hexagon pattern. Uneven Multi-Hex: performs a very wide search using a variety of patterns, more accurately capturing complex motion. Exhaustive: performs a \"dumb\" search of every pixel in a wide area. Significantly slower for only a small compression gain. Transformed Exhaustive: Like exhaustive, but makes even more accurate decisions. Accordingly, somewhat slower, also for only a small improvement. H.264 allows for two different prediction modes, spatial and temporal, in B-frames. Spatial, the default, is almost always better, but temporal is sometimes useful too. x264 can, at the cost of a small amount of speed (and accordingly for a small compression gain), adaptively select which is better for each particular frame. Sane values are ~2-5. This specifies the maximum number of sequential B-frames that the encoder can use. Large numbers generally won't help significantly unless Adaptive B-frames is set to Optimal. Cel-animated source material and B-pyramid also significantly increase the usefulness of larger values. Baseline profile, as required for iPods and similar devices, requires B-frames to be set to 0 (off). Sane values are ~1-6. The more you add, the better the compression, but the slower the encode. Cel animation tends to benefit from more reference frames a lot more than film content. Note that many hardware devices have limitations on the number of supported reference frames, so if you're encoding for a handheld or standalone player, don't touch this unless you're absolutely sure you know what you're doing! Performs extra analysis to decide upon weighting parameters for each frame. This improves overall compression slightly and improves the quality of fades greatly. Baseline profile, as required for iPods and similar devices, requires weighted P-frame prediction to be disabled. Note that some devices and players, even those that support Main Profile, may have problems with Weighted P-frame prediction: the Apple TV is completely incompatible with it, for example. B-pyramid improves compression by creating a pyramidal structure (hence the name) of B-frames, allowing B-frames to reference each other to improve compression. Requires Max B-frames greater than 1; optimal adaptive B-frames is strongly recommended for full compression benefit.