Original build
Completed in 2023 as a Jupyter notebook using Python, NumPy, OpenCV, scikit-image, SciPy, and Matplotlib, before generative AI coding tools were part of my workflow.
Open the notebook ↗Project 01
Image Signal Processor
A camera sensor records one red, green, or blue value at each pixel. This project reconstructs the missing color information and turns the raw Bayer frame into a finished image, one inspectable stage at a time.
- Completed
- 2023
- Original
- Python notebook
- Input
- 4096 × 2160 RGGB
- Demo
- Vanilla JavaScript
Live demo
Follow the image through the pipeline.
The page downloads the original 8.4 MB raw frame and processes it locally. A Web Worker is a separate browser thread, so the image math can run without freezing the controls or the rest of the page.
Loading the raw sensor data...
Raw Bayer data
The sensor records one color value at each pixel location.
- Original notebook
- NumPy fromfile + reshape
- Browser port
- Fetch + Uint8Array
Project notes
Implementation
Browser port
The static page fetches the 8.4 MB raw frame, performs Bayer-aware downsampling, and runs each stage on typed arrays in a Web Worker so the interface stays responsive.
What it demonstrates
The project covers raw image representation, color reconstruction, filter tradeoffs, and translating a notebook experiment into a performant interactive tool.