ICON: Learning Regular Maps Through Inverse Consistency
This addresses the challenge of ensuring well-behaved maps in applications like image registration, offering a simpler alternative to classical regularizers, though it appears incremental as it builds on existing deep learning approaches.
The authors tackled the problem of learning regular spatial maps, such as image registrations, by exploring if inverse consistency alone can induce regularity without explicit regularizers. They found that deep networks with an inverse consistency loss and randomized off-grid interpolation yield approximately diffeomorphic transformations, achieving competitive registration performance on synthetic and real data.
Learning maps between data samples is fundamental. Applications range from representation learning, image translation and generative modeling, to the estimation of spatial deformations. Such maps relate feature vectors, or map between feature spaces. Well-behaved maps should be regular, which can be imposed explicitly or may emanate from the data itself. We explore what induces regularity for spatial transformations, e.g., when computing image registrations. Classical optimization-based models compute maps between pairs of samples and rely on an appropriate regularizer for well-posedness. Recent deep learning approaches have attempted to avoid using such regularizers altogether by relying on the sample population instead. We explore if it is possible to obtain spatial regularity using an inverse consistency loss only and elucidate what explains map regularity in such a context. We find that deep networks combined with an inverse consistency loss and randomized off-grid interpolation yield well behaved, approximately diffeomorphic, spatial transformations. Despite the simplicity of this approach, our experiments present compelling evidence, on both synthetic and real data, that regular maps can be obtained without carefully tuned explicit regularizers, while achieving competitive registration performance.