DeepSim: Semantic similarity metrics for learned image registration
This addresses the challenge of aligning images with low intensity contrast or noise, which is incremental as it builds on existing learning-based registration models.
The authors tackled the problem of image registration by proposing a semantic similarity metric that learns dataset-specific features, achieving consistently high registration accuracy and faster convergence than state-of-the-art methods across multiple image modalities.
We propose a semantic similarity metric for image registration. Existing metrics like euclidean distance or normalized cross-correlation focus on aligning intensity values, giving difficulties with low intensity contrast or noise. Our semantic approach learns dataset-specific features that drive the optimization of a learning-based registration model. Comparing to existing unsupervised and supervised methods across multiple image modalities and applications, we achieve consistently high registration accuracy and faster convergence than state of the art, and the learned invariance to noise gives smoother transformations on low-quality images.