This research paper explores methods to compress facial images to under 1024 bytes for use in temporary travel documents, enabling machine readability via 2D barcodes. The study evaluates various compression formats including JPEG AI, AVIF, and WebP, alongside preprocessing techniques like grayscale conversion, smoothing, and resizing. JPEG AI, with optimized settings, demonstrated the best face recognition performance for high-quality images, while AVIF and WebP also yielded strong results. The findings suggest that significant compression is possible with minimal loss in recognition accuracy. AI
IMPACT Optimized compression could enable wider use of biometric verification in portable documents.
RANK_REASON The item is a research paper detailing image compression techniques for a specific application. [lever_c_demoted from research: ic=1 ai=0.7]
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
