Image registration method and device based on mitochondrion structure, storage medium and equipment

Abstract

The invention relates to the field of image processing and neurosciences, specifically relates to an image registration method and device based on mitochondrion structure, a storage medium and equipment and aims at solving a technical problem that electron microscopic section image registration precision of nervous tissue is low. For this purpose, the method comprises the steps: electron microscopic serial section images of a nervous tissue is acquired, mitochondrial region detection is performed on the acquired electron microscopic serial section image, according to detected feature points of the mitochondrial region, correspondences between the section images are matched, according to the matched correspondences, image deforming is performed on the electron microscopic serial section images, and the image registration is completed. In addition, the image registration device, the storage medium and equipment in the invention can execute steps in the above image registration method. According to the technical scheme of the invention, the precision of the nervous tissue electron microscopic serial section image registration can be improved, and mismatching in the image registration process can be avoided.

Description

technical field [0001] The invention relates to the fields of image processing and neuroscience, in particular to an image registration method and device, storage medium and device based on mitochondrial structure. Background technique [0002] With more and more countries and organizations launching brain science research programs, great progress has been made in the construction of brain maps at the macroscopic level at the millimeter scale and at the mesoscopic level at the micrometer scale. The progress of research at the nano-scale and micro-level is slightly slower, but in recent years, the work has also been carried out and the results have begun to show. In order to understand the relationship between the physical structure and function of neural circuits, the three-dimensional reconstruction of neuronal morphology and connection structure at the micro- and nano-scale is of great significance. This work mainly includes steps such as neural tissue sections, electron m...

AI Technical Summary

Problems solved by technology

However, since the images of nerve tissue slices contain a large number of similar synaptic structures and subcellular structures (such as mitochondria, vesicles, etc.), these similar structures will cause a large number of SIFT feature mismatches, and the RANSAC method cannot fully filter out the mismatched ones. Feature point pairs, which in turn affect the final image registration accuracy

Images