Robust image registration method based on association saliency image in global abnormal signal environment

Abstract

A robustness image registration method under a global abnormal signal circumstance and based on a combinated markedness figure comprises steps of: (1) firstly performing a cursory registration by a traditional registration method; (2) extracting a makedness measure figure from a current registration result and calculating the combinated makedness figure corresponding to an overlapping region, wherein combinated makedness figure provides a modeling at positions of abnormal signal points, increases contributed weight of a common markedness region in a similarity measure so that self-adaptively removing effect of the abnormal signals; (3) obtaining an optimum geometrical registration parameter of the current combinated markedness figure by using the combinated markedness figure to optimize the similarity measure based on gray scales; (4) cycling the optimization, finishing the registration by using variety between two registration parameters as a finish condition, thereby obtaining a final registration result. The invention iteratively performs the modeling of the abnormal signals and the registration for images, is suitable for multi-model multi-period image registration in the global abnormal signal circumstance, the registration precision is accorded with actual application.

Description

technical field [0001] The invention relates to a method in the technical field of image processing, in particular to a robust image registration method based on a joint saliency map in a global abnormal signal environment. Background technique [0002] Image registration is a basic and important work in computer vision, pattern recognition, image analysis and other application fields. It optimizes the image similarity measurement objective function and then searches for the optimal geometric transformation parameters for image matching, so that multiple images can be eliminated. The differences in space and time are integrated into a unified coordinate frame, which facilitates the complementary understanding of image information. Registration of images acquired in different modes and at different times is very important in medical imaging applications, remote sensing, computer vision, and defense imaging. Due to differences in physical characteristics of different imaging ...

AI Technical Summary

Problems solved by technology

The existence of abnormal signals will seriously affect the uniformity of the gray level, resulting in the uncertainty of the gray level correspondence between the two images. What is more serious is that this effect is often not spatially invariant.

The second type of method, that is, feature-based image registration method, although it has advantages in speed and efficiency in the process of extracting features and finding feature correspondences, it is difficult to meet the requirements of both rejecting the influence of global abnormal signals and effectively reconstructing Dual requirements for feature correspondences

"There is a special treatment for excluding overflow values ​​(abnormal), but this method does not consider the global abnormal signal that causes the lack of image feature correspondence, so it does not make special optimization for the global abnormal signal. In the case of a large number of abnormalities, its Registration results are not known

Images