Virtual ISP

A hands-on project to build a simplified Image Signal Processing (ISP) pipeline from scratch using only NumPy for pixel manipulation. This project avoids high-level OpenCV functions to provide a deeper understanding of the underlying image processing steps.

Project Workflow

1. Find a RAW image:
   • Download a .dng or .arw file (common in digital photography).
2. Demosaicing:
   • Implement an algorithm to convert the Bayer pattern (RGGB) into a full RGB image using NumPy.
3. White Balance:
   • Apply white balance (e.g., for 5500K daylight) by scaling the Red and Blue channels to neutralize color casts.
4. Gamma Correction:
   • Apply a non-linear gamma curve to the data for natural-looking output.

Getting Started

• Python 3.x
• NumPy

Goals

• Learn the fundamentals of digital image processing.
• Gain hands-on experience with RAW image data.
• Understand and implement demosaicing, white balance, and gamma correction from scratch.

License

MIT License