-
Notifications
You must be signed in to change notification settings - Fork 228
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
Showing
1 changed file
with
41 additions
and
20 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -1,28 +1,49 @@ | ||
# Improved JPEG decoder implementation | ||
# Improved JPEG encoder and decoder implementation | ||
|
||
This subdirectory contains a JPEG decoder implementation that is API and ABI | ||
compatible with libjpeg62. | ||
This subdirectory contains a JPEG encoder and decoder implementation that is | ||
API and ABI compatible with libjpeg62. | ||
|
||
*NOTE*: This is still a work in progress, currently only API functions called | ||
from libjxl's benchmark_xl tool are implemented. | ||
## Building | ||
|
||
To decompress an ```input.jpg``` file with this new library: | ||
When building the parent libjxl project, two binaries, `tools/cjpegli` and | ||
`tools/djpegli` will be built, as well as a | ||
`lib/jpegli/libjpeg.so.62.3.0` shared library that can be used as a drop-in | ||
replacement for the system library with the same name. | ||
|
||
``` | ||
(from the libjxl root directory) | ||
$ ./ci.sh opt | ||
$ LD_PRELOAD=./build/libjpeg.so.62 ./build/tools/benchmark_xl --input input.jpg --codec=jpeg --decode_only --save_decompressed --output_dir . | ||
``` | ||
## Encoder improvements | ||
|
||
The decompressed file will be saved as ```input.jpg.jpeg.png```. | ||
Improvements and new features used by the encoder include: | ||
|
||
To benchmark the jpeg encoding-decoding round-trip on an ```input.png``` with | ||
the new library, first build a statically linked ```cjpeg-static``` binary, | ||
which is found in ```$PATH```, and then run: | ||
* Support for 16-bit unsigned and 32-bit floating point input buffers. | ||
|
||
``` | ||
(from the libjxl root directory) | ||
$ ./ci.sh opt | ||
$ LD_PRELOAD=./build/libjpeg.so.62 ./build/tools/benchmark_xl --input input.png --codec=jpeg:cjpeg-static:q90 | ||
``` | ||
* Color space conversions, chroma subsampling and DCT are all done in floating | ||
point precision, the conversion to integers happens first when producing | ||
the final quantized DCT coefficients. | ||
|
||
* The desired quality can be indicated by a distance parameter that is | ||
analogous to the distance parameter of JPEG XL. The quantization tables | ||
are chosen based on the distance and the chroma subsampling mode, with | ||
different positions in the quantization matrix scaling differently, and the | ||
red and blue chrominance channels have separate quantization tables. | ||
|
||
* Adaptive dead-zone quantization. On noisy parts of the image, quantization | ||
thresholds for zero coefficients are higher than on smoother parts of the | ||
image. | ||
|
||
* Support for more efficient compression of JPEGs with an ICC profile | ||
representing the XYB colorspace. These JPEGs will not be converted to the | ||
YCbCr colorspace, but specialized quantization tables will be chosen for | ||
the original X, Y, B channels. | ||
|
||
## Decoder improvements | ||
|
||
* Support for 16-bit unsigned and 32-bit floating point output buffers. | ||
|
||
* Non-zero DCT coefficients are dequantized to the expectation value of their | ||
respective quantization intervals assuming a Laplacian distribution of the | ||
original unquantized DCT coefficients. | ||
|
||
* After dequantization, inverse DCT, chroma upsampling and color space | ||
conversions are all done in floating point precision, the conversion to | ||
integer samples happens only in the final output phase (unless output to | ||
floating point was requested). |