This notebook explores the mechanisms and algorithms at play behind one of the most widely used image file formats. By parsing the file byte by byte, and using a specification reference, we can interpret all of the information stored in the file, and generate the corresponding image without a pre-built image viewer. The primary reference used will be the International Telecommunication Union Recommendation T.81.
All functions are written using Python with diagrams and full explanations to accompany them. The following are some of the decoding steps discussed in the notebook:
- Segment Parsing
- Huffman Decoding
- Quantization
- MCU Building
With Jupyter installed, download jpeg_decoding.ipynb, and in its directory run
> jupyter notebook jpeg_decoding.ipynbThe project is also available in HTML format.