Tiny object segmentation method and device based on super-resolution technology

Abstract

The invention discloses a subtle object segmentation method and device based on a super-resolution technology. The subtle object segmentation method and device can be used for segmentation of three-dimensional tiny blood vessels. The method comprises the following steps: 1) constructing a super-resolution network, providing a structure-maintained super-resolution network, and carrying out super-resolution operation on tiny blood vessels on the basis of maintaining blood vessel structure information to obtain more detail information; (2) a three-dimensional blood vessel segmentation network is constructed, the segmentation network adopts a U-shaped network structure, two layers which are symmetrical end to end are linked while features are extracted layer by layer in a convolutional layer, and sufficient extraction of the features is achieved; and 3) processing the data by using super-resolution, and training the segmentation network by using the data after super-resolution, thereby achieving a better segmentation effect. According to the method, the blood vessel characteristic information is enhanced through the super-resolution technology, so that the original network is obviously improved, and the efficient and accurate segmentation effect is realized.

Description

technical field [0001] The invention belongs to the technical field of image segmentation in computer vision, and in particular relates to a method and device for segmenting fine objects based on super-resolution technology. Background technique [0002] Image segmentation technology is an important research direction in the field of computer vision and an important part of image semantic understanding. Image segmentation is a technique and process of dividing an image into several specific regions with unique properties and proposing objects of interest. It has broad application prospects in the medical field. [0003] Because convolutional neural network has excellent feature extraction ability and good feature expression ability, it does not need to manually extract image features or perform too much image preprocessing, so in recent years, convolutional neural network has achieved great success in the field of medical image segmentation. Achievements, such as the segmen...

AI Technical Summary

Problems solved by technology

At the same time, the super-resolution network with the gradient branch added, compared with the traditional super-resolution algorithm, focuses on structural features, which can effectively solve the problem of structural distortion in traditional super-resolution images, and provide convenience for subsequent segmentation work

Images