BmpVsJpg

Educational JPEG/BMP image toolkit. Includes a portable C++ core library, a cross-platform CLI tool, and a Windows MFC GUI. Based on the book "Visual C++ MPEG-JPEG".

JPEG/BMP 教学工具集。包含跨平台 C++ 核心库、命令行工具和 Windows MFC GUI。

Install from Release

Download prebuilt binaries from the Releases page:

# macOS (Apple Silicon)
curl -LO https://github.com/XWHQSJ/BmpVsJpg/releases/latest/download/jpeg_edu-macos-arm64-v1.0.0.tar.gz
tar xzf jpeg_edu-macos-arm64-v1.0.0.tar.gz

# Linux (x86_64)
curl -LO https://github.com/XWHQSJ/BmpVsJpg/releases/latest/download/jpeg_edu-linux-x86_64-v1.0.0.tar.gz
tar xzf jpeg_edu-linux-x86_64-v1.0.0.tar.gz

Requires libjpeg-turbo at runtime (brew install jpeg-turbo / apt install libjpeg-turbo8).

Three Consumption Modes

1. CLI Tool (jpeg_edu)

# Build
cmake -B build -S .
cmake --build build

# Encode BMP -> JPEG
./build/jpeg_edu encode photo.bmp photo.jpg -q 85 -progressive

# Decode JPEG -> BMP
./build/jpeg_edu decode photo.jpg output.bmp

# Image info
./build/jpeg_edu info photo.jpg

# Apply filter
./build/jpeg_edu filter input.bmp edges.bmp -f sobel

# Quantize colors (octree)
./build/jpeg_edu quantize input.bmp reduced.bmp -n 32

# Benchmark encode/decode
./build/jpeg_edu bench input.bmp -q 85 -rounds 10

2. C++ Library (libjpeg_edu)

#include "libjpeg_edu/decoder.h"
#include "libjpeg_edu/encoder.h"
#include "libjpeg_edu/filters.h"

using namespace libjpeg_edu;

auto img = decode_jpeg("photo.jpg");
auto edges = apply_filter(img, FilterType::kSobel);
encode_jpeg("edges.jpg", edges, {.quality = 90, .progressive = true});

Link against libjpeg_edu in your CMakeLists.txt:

add_subdirectory(path/to/BmpVsJpg)
target_link_libraries(your_target PRIVATE libjpeg_edu)

3. MFC GUI (Windows only)

Open BmpVsJpg/BmpVsJpg.sln in Visual Studio with MFC support. The GUI provides BMP-to-JPEG and JPEG-to-BMP conversion with quality control and image processing filters.

Build Requirements

Platform	Requirements
macOS/Linux	CMake 3.14+, C++17, libjpeg or libjpeg-turbo
Windows	Visual Studio 2019+, CMake, or direct .sln for MFC

# macOS
brew install jpeg-turbo cmake
cmake -B build -S . && cmake --build build

# Run tests
ctest --test-dir build --output-on-failure

Project Structure

BmpVsJpg/
  libjpeg_edu/             Portable C++ core library (no Windows deps)
    include/libjpeg_edu/   Public headers
    src/                   Implementation
  cli/                     Cross-platform CLI tool
  tests/                   GoogleTest roundtrip + PSNR tests
  fuzz/                    libFuzzer targets + seed corpus
  docs/                    Quality comparison + filter showcase HTML
  qa/                      Asset generation scripts
  .github/workflows/       CI (sanitizers + fuzz + CodeQL + release)
  BmpVsJpg/                Original MFC GUI (Windows only)
  Change_Bmp_Jpg/          Alternative MFC implementation
  CMakeLists.txt           Top-level build

Library Modules

Module	Description
decoder.h	JPEG decode via libjpeg, metadata query
encoder.h	JPEG encode with quality, grayscale, progressive
bmp_io.h	24-bit BMP read/write with BGR/bottom-up handling
ppm_io.h	PPM (P6) and PGM (P5) read/write
filters.h	Blur, sharpen, edge, Sobel, median convolution filters
octree.h	Octree color quantization

Tests

14 GoogleTest cases covering JPEG roundtrip PSNR (>35 dB @ q95, >25 dB @ q50), BMP/PPM lossless roundtrip, filter output, octree quantization, and error handling.

CI runs all tests under AddressSanitizer and UndefinedBehaviorSanitizer on both Ubuntu and macOS.

Visual Demos

• Quality Comparison -- the same image at quality 10/25/50/75/95/100 with file sizes and PSNR
• Filter Showcase -- blur, sharpen, edge, Sobel, and median filter effects with kernel visualizations

Regenerate assets: qa/generate_comparison.sh (requires a build of jpeg_edu).

Fuzzing

libFuzzer targets for the JPEG decoder and BMP reader live in fuzz/. CI runs each for 5 minutes per push with ASan+UBSan enabled.

# Build and run locally (requires LLVM clang, not Apple clang)
cmake -B build-fuzz -DENABLE_FUZZ=ON \
  -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ .
cmake --build build-fuzz --target fuzz_decoder
./build-fuzz/fuzz/fuzz_decoder fuzz/corpus/ -max_total_time=60

Seed corpus in fuzz/corpus/ contains JPEGs at various quality levels plus a BMP.

Cleanup Notes

The duplicate BmpToJpg/ and JpgToBmp/ directories have been removed (they were identical copies). Bug fixes applied to Jpeg.cpp (tb1/tb2 pointer check) and BmpVsJpgDlg.cpp (hardcoded quality).

References

• Reference book: Visual C++ MPEG-JPEG
• Book PDF and source code: Baidu Netdisk password: dgpt
• Additional source code and libraries: Baidu Netdisk password: yw2v

License

MIT

Security

See SECURITY.md for vulnerability reporting and security measures.