This is a C++ Library which implements a number of algorithms that are used for compression of multimedia objects.

I took a course in my sixth semester called "Multimedia Systems and Applications" which motivated me to start this project. The algorithms implemented in this repository are:

• Lossless Compression Algorithms
  • Huffman Coding
  • Lempel-Ziv Coding
  • Run-Length Encoding and BitPlane conversion
  • Lossless Predictive Coding
• Lossy Compression Algorithms
  • Lossy Predictive Coding
  • Differential Pulse Code Modulation (in lossy/lossy_pred)
• Transform-Domain Coding
  • K-L Transform
  • Discrete Cosine Transform (DCT) and its inverse
  • Bit allocation methods for DCT (Zonal and Threshold coding)
• Wavelets and MultiResolution Analysis
  • Subband coding with 4 bands and reconstruction
  • And by extension, the DWT since it can be computed by changing the filters in the subband coding architecture
  • Embedded Zerotree Wavelet (EZW) coding

We use BitMap (.bmp) image files for testing the algorithms because they are in the most uncompressed form.

1. Clone the repository and cd into it.
2. Create the CMake build files in a directory called build by running the command cmake -B build -G "MinGW Makefiles" -DOpenCV_DIR=<YOUR_OPENCV_DIRECTORY>. The placeholder YOUR_OPENCV_DIRECTORY should contain the location of your opencv installation, for me it was C:/opencv/build since I installed it manually.
3. cd into the build folder. Build using cmake --build .
4. Run the executable.

1. The compression achieved via lossless predictive coding, combined with Huffman encoding of the error values, is extremely substantial, with a randomly chosen coefficient set of [0, 1, 3, 2] (before normalization) yielding 81% compression for the grayscale image grayscale.bmp, as compared to 8% with only Huffman encoding.
2. Quantization improves the compression ratio, although with the tradeoff of image quality. With just 4 levels of quantization, we can get upto 93% compression, although with significantly visible deterioration of quality.