Real-Time Demosaicking for Embedded Systems
Wei Tse Hsu, Chiou‐Shann Fuh · 2007
We present an effective and efficient demosaicking algorithm for natural images. A new reciprocal table accurately estimates the relation between edges and gradient weights, and it achieves regular, succinct, and fast operations with excellent performance. Flexible and optional procedures adapt to different constraints on embedded systems. Our proposed method has 10.94 dB, 9.14 dB, and 9.05 dB improvements over bilinear method in red, green, and blue channels respectively. In visual results, the reconstructed images have smooth artifact. The performance is very close to recent methods, still superiorly.