PCReg-Net: Progressive Contrast-Guided Registration for Cross-Domain Image Alignment

Researchers have developed PCReg-Net, a lightweight AI framework for cross-domain image registration that achieves real-time performance at 141 FPS with only 2.56M parameters. The system uses a progressive contrast-guided approach with four modules to align images across different domains, showing improvements over traditional and deep learning baselines on retinal and microscopy benchmarks.

• →PCReg-Net introduces a novel progressive contrast-guided framework for deformable image registration across heterogeneous domains.
• →The system achieves real-time inference at 141 FPS while using only 2.56M parameters, making it highly efficient.
• →Four lightweight modules work together to perform coarse-to-fine alignment: registration U-Net, feature extractor, contrast module, and refinement U-Net.
• →Testing on FIRE-Reg-256 retinal fundus and microscopy benchmarks demonstrates superior performance over existing methods.
• →The framework addresses brightness constancy assumption violations that challenge conventional registration methods.