Scarfski/basis_universal
1
0
Fork 0
mirror of https://github.com/BinomialLLC/basis_universal.git synced 2026-09-10 12:38:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update README.md

Browse Source
This commit is contained in:
Rich Geldreich
2019-05-18 18:12:36 -07:00
committed by GitHub
parent 9529bdbff8
commit ebd77b8282
1 changed files with 8 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
README.md
View File

@@ -1,9 +1,9 @@
# basis_universal
Basis Universal GPU Texture Compression Codec
Basis Universal GPU Texture/Video Compression Reference Encoder
Basis Universal is a ["supercompressed"](http://gamma.cs.unc.edu/GST/gst.pdf) GPU texture compression system that outputs a highly compressed intermediate file format (.basis) that can be quickly transcoded to a wide variety of GPU texture compression formats: PVRTC1 4bpp RGB, BC7 mode 6 RGB, BC1-5, ETC1, and ETC2. We will be adding ASTC RGB or RGBA, BC7 mode 4/5 RGBA, and PVRTC1 4bpp RGBA next. Basis files support non-uniform texture arrays, so cubemaps, volume textures, texture arrays, mipmap levels, video sequences, or arbitrary texture "tiles" can be stored in a single file. The compressor is able to exploit color and pattern correlations across the entire file, so multiple images with mipmaps can be stored very efficiently in a single file.
Basis Universal is a ["supercompressed"](http://gamma.cs.unc.edu/GST/gst.pdf) GPU texture and video compression system that outputs a highly compressed intermediate file format (.basis) that can be quickly transcoded to a wide variety of GPU texture compression formats: PVRTC1 4bpp RGB, BC7 mode 6 RGB, BC1-5, ETC1, and ETC2. We will be adding ASTC RGB or RGBA, BC7 mode 4/5 RGBA, and PVRTC1 4bpp RGBA next. Basis files support non-uniform texture arrays, so cubemaps, volume textures, texture arrays, mipmap levels, video sequences with I-Frames and P-Frames using Conditional Replenishment (CR), or arbitrary texture "tiles" can be stored in a single file. The compressor is able to exploit color and pattern correlations across the entire file, so multiple images with mipmaps can be stored very efficiently in a single file. For video, the system is able to entirely skip blocks which don't change from the previous frame, which can be very effective.
The system's bitrate depends on the quality setting and image content, but common usable bitrates are .3-1.25 bits/texel. .basis files are typically 10-25% smaller than using RDO texture compression of the internal texture data stored in the .basis file followed by LZMA. The current system is what we've been calling the "baseline" system, which is designed to reach all the GPU formats. The next major step is to extend the system to allow for much higher quality for the ASTC and BC7 texture formats.
The system's bitrate depends on the quality setting and image content, but common usable bitrates are .3-1.25 bits/texel. .basis files are typically 10-25% smaller than using RDO texture compression of the internal texture data stored in the .basis file followed by LZMA. For video, the average bitrate will highly depend on how dynamic the content is, but we usually get around .3-.5 bits/texel on average across long video sequences. The current system is what we've been calling the "baseline" system, which is designed to reach all the GPU formats. The next major step is to extend the system to allow for much higher quality for the ASTC and BC7 texture formats.
The transcoder has been fuzz tested using [zzuf](https://www.linux.com/news/fuzz-testing-zzuf).
@@ -61,11 +61,13 @@ Note that "-no_selector_rdo -no_endpoint_rdo" are optional. Using them hurts rat
To compress small video sequences, say using tools like ffmpeg and VirtualDub:
'basisu -slower -tex_type video -stats -debug -multifile_printf "pic%04u.png" -multifile_num 200 -multifile_first 1 -max_selectors 16128 -max_endpoints 16128'
'basisu -slower -tex_type video -framerate 25 -stats -debug -multifile_printf "pic%04u.png" -multifile_num 200 -multifile_first 1 -max_selectors 16128 -max_endpoints 16128'
The reference encoder will take a LONG time and a lot of CPU to encode video, especially with -slower. The more cores your machine has, the better. Basis is intended for smaller videos of a few dozen seconds or so. If you are very patient and have a Threadripper or Xeon workstation, you should be able to encode up to a few thousand 720P frames.
The reference encoder will take a LONG time and a lot of CPU to encode video, especially with -slower. (-slower is needed to generate better codebooks, which many videos need.) The more cores your machine has, the better. Basis is intended for smaller videos of a few dozen seconds or so. If you are very patient and have a Threadripper or Xeon workstation, you should be able to encode up to a few thousand 720P frames. Over time, we will be optimizing the codebook generators for higher performance, especially with video.
The .basis file will contain multiple images (all using the same global codebooks), which you can retrieve using the transcoder's image API. This initial release doesn't support [conditional replenisment](https://en.wikipedia.org/wiki/MPEG-1) (CR), but we have a branch in the works that does that doesn't change the file format itself. CR can reduce the bitrate of some videos (highly dependent on how dynamic the content is) by over 50%. For videos using CR, the images must be requested from the transcoder in sequence from first to last, and random access is only allowed to I-Frames. (More on this once we release it.)
The .basis file will contain multiple images (all using the same global codebooks), which you can retrieve using the transcoder's image API. For videos, the images must be requested from the transcoder in sequence from first to last, and random access is only allowed to I-Frames. Currently, the first image is always an I-Frame, and all subsequent images are P-Frames that can use CR. Videos can optionally contain mipmaps and alpha channels.
There's a very simple video playback demo in the "webgl_videotest" directory. Note this example currently uses asm.js, not WebAssembly, but video is fully compatible with WebAssembly.
If you are doing rate distortion comparisons vs. other similar systems, be sure to experiment with increasing the endpoint RDO threshold (-endpoint_rdo_thresh X). This setting controls how aggressively the compressor's backend will combine together nearby blocks so they use the same block endpoint codebook vectors, for better coding efficiency. X defaults to a modest 1.5, which means the backend is allowed to increase the overall color distance by 1.5x while searching for merge candidates. The higher this setting, the better the compression, with the tradeoff of more block artifacts. Settings up to ~2.25 can work well, and make the codec more competitive.