Home

Awesome

jpeg-compressor

A small (~1000 lines), easy to use public domain (or Apache 2.0) C++ class in a single source file jpge.cpp that writes baseline JPEG compressed images. It supports grayscale and H1V1/H2V1/H2V2 chroma subsampling factors, Libjpeg-compatible quality settings, and is reasonably fast with fairly low (typically less than 64KB) memory consumption. The core compression class consists of a single 890 line C++ file with a small header, along with a couple optional higher-level helper/example functions. The current release supports both single pass Huffman coding and more efficient (but slower) two pass coding, makes only a single dynamic memory allocation, and now accepts 32-bit source images.

The source distribution also includes an optional, completely stand-alone public domain (or Apache 2.0) JPEG decompression class with progressive image support in a single source file jpgd.cpp. It supports both box and linear chroma upsampling, and grayscale or H1V1/H2V1/H1V2/H2V2 chroma upsampling factors. Unlike every other small JPEG decompressor I've seen, this decompressor has been fuzz tested using zzuf and afl, making it resilent against crashing, overwriting memory, or bad memory reads when given accidently or purposely corrupted inputs. Also unlike many other small JPEG decompressors, jpgd.cpp does not require loading the entire image into memory, just single MCU rows at a time (even when doing linear chroma upsampling). This property makes the decompressor useful on small 32-bit microcontrollers.

The source distribution includes a sample VS2019 Win32/x64 solution and CMakeLists.txt file for compilation with gcc/clang.

Thanks to Alex Evans for adding several features to jpge (see a smaller jpg encoder).

Basic Usage (Compression)

Include jpge.h and call one of these helper functions in the "jpge" namespace:

// Writes JPEG image to a file. 
// num_channels must be 1 (Y), 3 (RGB), or 4 (RGBA), image pitch must be width*num_channels.
bool compress_image_to_jpeg_file(const char *pFilename, int width, int height, int num_channels, 
                                 const uint8 *pImage_data, const params &comp_params = params());

// Writes JPEG image to memory buffer. 
// On entry, buf_size is the size of the output buffer pointed at by pBuf, which should be at least ~1024 bytes. 
// If return value is true, buf_size will be set to the size of the compressed data.
bool compress_image_to_jpeg_file_in_memory(void *pBuf, int &buf_size, int width, int height, int num_channels, 
                                           const uint8 *pImage_data, const params &comp_params = params());

See tga2jpg.cpp for an example usage. This example uses Sean Barrett's stb_image.c module to load image files.

You can also call the jpge::jpeg_encoder class directly if you need more control over the image source or how/where the output stream is written.

Basic Usage (Decompression)

Include jpgd.h and call one of these helper functions in the "jpgd" namespace:

// Loads a JPEG image from a memory buffer.
// req_comps can be 1 (grayscale), 3 (RGB), or 4 (RGBA).
// On return, width/height will be set to the image's dimensions, and actual_comps will be set 
// to either 1 (grayscale) or 3 (RGB).
unsigned char *decompress_jpeg_image_from_memory(const unsigned char *pSrc_data, int src_data_size, 
                                     int *width, int *height, int *actual_comps, int req_comps, uint32_t flags = 0);

// Loads a JPEG image from a file.
unsigned char *decompress_jpeg_image_from_file(const char *pSrc_filename, int *width, int *height, 
                                     int *actual_comps, int req_comps, uint32_t flags = 0);

Just like the compressor, for more control you can directly utilize the jpgd::jpeg_decompressor class, or call the decompress_jpeg_image_from_stream() function.

The optional flags are:

cFlagBoxChromaFiltering: Uses much faster box filtering on the chroma components, but may lead to artifacts on some images

cFlagDisableSIMD: Do not use SSE2, even if it's been compiled in

Set the "JPGD_USE_SSE2" macro to 0 to completely disable SSE2 usage during compilation. By default, it's enabled (unless it's x86/x64 gcc and SSE2 is not defined).

jpgd supports a superset of JPEG features utilized by jpge, so it can decompress any file generated by jpge along with most (if not almost all) JPEG files you're likely to encounter in the wild. It supports progressive and baseline sequential JPEG image files and the H1V1, H2V1, H1V2, and H2V2 chroma subsampling factors.

Testing

The source distribution includes a simple example command line tool called "jpge.exe" (or jpge_x64.exe for Win64 systems) that converts images from any format that stb_image.c supports (such as PNG, TGA, BMP, etc.) to baseline JPEG using the jpeg_encoder class. Its usage is: bin/jpge source_file.png destination_file.jpg quality_factor Where quality_factor ranges from 1-100 (higher is better).

jpge.exe also includes a few other modes for exhaustive testing of the codec (-x option) and JPEG to TGA decompression (-d option) -- see the help text for more info.

Licenses

The license of jpgd.cpp/.h and jpge.cpp/.h is either Public Domain or Apache 2.0. Choose whatever you want.

The license for the optional file jpgd_idct.h (and ONLY this file) is Copyright 2009 Intel Corporation:

Permission is granted to use, copy, distribute and prepare derivative works of this software for any purpose and without fee, provided, that the above copyright notice and this statement appear in all copies. Intel makes no representations about the suitability of this software for any purpose. THIS SOFTWARE IS PROVIDED "AS IS." INTEL SPECIFICALLY DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, AND ALL LIABILITY, INCLUDING CONSEQUENTIAL AND OTHER INDIRECT DAMAGES, FOR THE USE OF THIS SOFTWARE, INCLUDING LIABILITY FOR INFRINGEMENT OF ANY PROPRIETARY RIGHTS, AND INCLUDING THE WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. Intel does not assume any responsibility for any errors which may appear in this software nor any responsibility to update it.

(Set JPGD_USE_SSE2 to 0 to completely disable usage of jpgd_idct.h.)

Revisions

History

This codec was originally written in C and 16-bit x86 asm way back in 1994 for a DOS image viewer. The primary reference was Pennebaker's and Mitchell's book JPEG: Still Image Data Compression Standard. The original version supported a simple form of adaptive quantization, used a quantized DCT, and two pass Huffman coding. Around 2000 I ported it to C++, but I didn't really have the time to release it until now.

Note if you're generating texture mipmaps from loaded images, you may be interested in my imageresampler project.

For any questions or problems with this code please contact Rich Geldreich at <richgel99 at gmail.com>. Here's my twitter page.