Feat: Add Title ROI masking

[poketonova]

Feat: Add Title ROI masking

Overview

This PR introduces robust handling for scanner noise and detection errors, ensuring high-fidelity card processing.

1. Title ROI Masking

Implements bitwise AND masking for the top 15% of the homography result to isolate the card title, effectively reducing scanner artifact noise. Unit testing for this logic is implemented via tests/scanner/ocr.spec.js using jest-image-snapshot. Note that this approach introduces a minor, calculated latency overhead per frame.

2. Lighting Compensation

Addresses variable ambient lighting conditions by applying Contrast Limited Adaptive Histogram Equalization (CLAHE) on the Y channel, coupled with Adaptive Otsu's thresholding implemented within a WebGL fragment shader for GPU-accelerated performance.

3. Mis-detection Handling Strategy

Implements a multi-layered verification strategy:

• Secondary OCR Validation: Hamming distance comparison against the master price list.
• Confidence Score Thresholding: Detections with confidence < 0.85 are automatically flagged for manual review.
• Geometric Validation: ORB feature matching as a geometry validation layer against card template features.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
pokemon-collectr	Ready	Preview, Comment	Jun 12, 2026 3:05pm