Image processing method and apparatus

Abstract

The invention discloses an image processing method and apparatus, and belongs to the field of image recognition. The method comprises the steps of acquiring a target image; calling an image recognition model, wherein the image recognition model comprises a backbone network, and a pooling module and a cavity convolution module connected with the backbone network and parallel to each other, and a fusion module connected with the pooling module and the cavity convolution module; performing feature extraction on the target image through the backbone network in the image recognition model, processing characteristic images output by the backbone network through the pooling module and the cavity convolution module to obtain a first result output by the pooling module and a second result output bythe cavity convolution module, fusing the first result and the second result through the fusion module, and outputting a model recognition result of the target image; and based on the model recognition result, obtaining a semantic segmentation label graph of the target image. By adopting the method and the apparatus, the demands of accuracy and segmentation precision can be met at the same time.

Description

technical field [0001] The invention relates to the field of image recognition, in particular to an image processing method and device. Background technique [0002] Convolutional neural network is a machine learning model widely used in the field of image processing. For the medical field, convolutional neural networks can be used as image recognition models, for example, to identify cancer tissue regions in gastric pathology images. [0003] Generally speaking, there are three main structures of convolutional neural networks used to recognize pathological images: (1) full convolutional networks; (2) convolutional neural networks based on dilated convolutions; (3) spatial pooling-based Convolutional neural network. The server can call the image recognition model to classify each pixel of the pathological image (such as cancer or no cancer), based on the output model recognition results, mark the pixels of different classifications, and obtain the semantic segmentation mar...

AI Technical Summary

Problems solved by technology

When the method of full convolutional network is applied to large and small cancer tissue regions, the recognition accuracy is low; the method based on hole convolution can only be used for cancers of a certain size range under a single dilation rate. The tissue region has a good segmentation effect, which cannot meet the size range of the actual cancer tissue region; the method based on spatial pooling can take into account both large-size and small-size cancer tissue regions, but it will cause a large loss of information in the pathological image, which will bring Loss of Segmentation Accuracy

The structure of any of the above convolutional neural networks has certain limitations for segmenting gastric pathological images

Images