Fast alignment of digital images using a lower bound on an entropy metric

We propose a registration algorithm based on successively refined quantization and an alignment metric derived from a minimal spanning tree entropy estimate. The metric favors edge alignment, is fast to compute, and compares well in experiments with competing approaches.

• Type: Conference
• Year: 2004
• Where: ICIP
• Authors: Mert R. Sabuncu, Peter J. Ramadge
• Keywords: Favors Edge Alignment, ICIP 2004, Image Processing