Wavec - Wavelet Image Codec

A simple wavelet-based image compression tool built through vibe coding experimental, exploratory, and made for fun.

What is this?

Wavec is a Windows desktop application that implements wavelet-based image compression. It can:

• Load 24-bit BMP images
• Apply 2D Discrete Wavelet Transform (DWT) using Haar, Daubechies-4, or CDF 9/7 wavelets
• Discard small coefficients for compression
• Quantize remaining coefficients
• Save to a custom .WT sparse format
• Reconstruct images using Inverse DWT

Important: Image Requirements

Use BMP images with dimensions that are powers of 2 (e.g., 256×256, 512×512, 1024×1024, 512×256, etc.)

This ensures optimal wavelet decomposition across all levels. Non-power-of-2 dimensions may work but could produce unexpected results or artifacts.

Building

Requirements

• Windows 10/11
• Visual Studio 2022 (Community edition works fine)
• Windows SDK

Compilation

Run the provided batch file:

compile.bat

Or manually:

call "C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Auxiliary\Build\vcvars64.bat"
rc wavec.rc
cl wavec.cpp wavec.res user32.lib gdi32.lib comdlg32.lib comctl32.lib /Fe:wavec.exe /O2

Usage

1. Open a BMP File > Open BMP... or Ctrl+O
2. Transform Process > Transform... or F5
   • Select wavelet type (Haar, Db4, CDF 9/7)
   • Choose decomposition levels (Auto recommended)
   • Set discard percentage (higher = more compression, more loss)
   • Adjust quantization bits (lower = smaller file, more artifacts)
3. Save as WT File > Save WT... to save compressed format
4. Save as BMP File > Save BMP... to export reconstructed image
5. Reset F7 to restore original image

Keyboard Shortcuts

Key	Action
F5	Transform
F7	Reset
Ctrl+O	Open BMP
Ctrl+S	Save BMP

The .WT Format

A custom sparse format that stores only non-zero wavelet coefficients as (index, quantized_value) pairs. Includes metadata for reconstruction:

• Image dimensions
• Wavelet type and decomposition levels
• Quantization parameters
• Coefficient value range for dequantization

Wavelets Implemented

Wavelet	Description
Haar	Simplest wavelet, uses averages and differences
Daubechies-4	4-tap orthogonal wavelet with better frequency localization
CDF 9/7	Biorthogonal wavelet used in JPEG2000, implemented via lifting scheme

Disclaimer

This is a vibe coding project built experimentally without rigorous testing or production-quality standards. It's meant for learning, exploration, and having fun with wavelet transforms. Use at your own risk!

References

• Haar Wavelet
• Daubechies Wavelets
• CDF 9/7 Wavelet
• Lifting Scheme

License

Do whatever you want with it.