Abstract:Registration of retinal images taken at different times, from different perspectives, or with different modalities is a critical prerequisite for the diagnoses and treatments of various eye diseases. This problem can be formulated as registration of two sets of sparse feature points extracted from the given images, and it is typically solved by first creating a set of putative correspondences and then removing the false matches as well as estimating the spatial transformation between the image pairs or solved by estimating the correspondence and transformation jointly involving an iteration process. However, the former strategy suffers from missing true correspondences, and the latter strategy does not make full use of local appearance information, which may be problematic for low-quality retinal images due to a lack of reliable features. In this paper, we propose a feature-guided Gaussian mixture model (GMM) to address these issues. We formulate point registration as the estimation of a feature-guided mixture of densities: A GMM is fitted to one point set, such that both the centers and local features of the Gaussian densities are constrained to coincide with the other point set. The problem is solved under a unified maximum-likelihood framework together with an iterative expectation-maximization algorithm initialized by the confident feature correspondences, where the image transformation is modeled by an affine function. Extensive experiments on various retinal images show the robustness of our approach, which consistently outperforms other state-of-the-art methods, especially when the data is badly degraded.

Feature guided Gaussian mixture model with semi-supervised EM and local geometric constraint for retinal image registration.

Retinal Image Registration Via Feature-Guided Gaussian Mixture Model.

Multimodal Retinal Image Registration Using Edge Map and Feature Guided Gaussian Mixture Model.

Robust Image Matching Via Feature Guided Gaussian Mixture Model

Feature-guided Gaussian Mixture Model for Image Matching.

Multiple Image Features-Based Retinal Image Registration Using Global and Local Geometric Structure Constraints.

Gaussian field estimator with manifold regularization for retinal image registration.

Removing Mismatches for Retinal Image Registration Via Multi-Attribute-driven Regularized Mixture Model.

Non-Rigid Point Set Registration Via Gaussians Mixture Model with Local Constraints

Retinal fundus image registration framework using Bayesian integration and asymmetric Gaussian mixture model

Non-Rigid Point Set Registration Using Multi-Feature and Gaussian Mixture Model

Robust Non-Rigid Feature Matching for Image Registration Using Geometry Preserving

An Accurate Registration Method Based on Global Mixed Structure Similarity (GMSIM) for Remote Sensing Images

OPTICAL-TO-SAR IMAGE REGISTRATION BASED ON GAUSSIAN MIXTURE MODEL

Fuzzy Correspondences Guided Gaussian Mixture Model for Point Set Registration.

Retinal Image Registration Based on Robust Non-Rigid Point Matching Method

Non-rigid point set registration based on local neighborhood information support

Global Optimization Surface-Based Registration for Image-to-Patient Registration Using Gaussian Mixture Model

Fast Rotation-Free Feature-Based Image Registration Using Improved N-SIFT and GMM-Based Parallel Optimization

Feature Guided Biased Gaussian Mixture Model for Image Matching

Retinal Fundus Image Registration Via Vascular Structure Graph Matching.