Rapid high-resolution image segmentation method based on block recommendation network

Abstract

The invention discloses a rapid high-resolution image segmentation method based on a block recommendation network, and relates to image processing. The method comprises the following steps: 1) constructing a global branch and a local refined branch; 2) performing down-sampling on the original high-resolution image, and uniformly dividing the original high-resolution image into a plurality of imageblocks; 3) inputting the down-sampled image into a global branch to obtain a global segmentation feature map, and uniformly dividing the global segmentation feature map into a plurality of feature blocks; 4) inputting the down-sampled image into a block recommendation network to obtain a recommendation block; 5) taking out the recommendation block according to the recommendation block label, performing significance operation on the recommendation block and the feature block corresponding to the global segmentation feature map, and inputting a result into a local refined branch; 6) fusing corresponding positions of the local refined feature blocks and the global segmentation feature map, and outputting a fused segmentation result as an overall segmentation result; 7) calculating error lossbetween the segmentation result and a real label, training a network and updating network parameters, and 8) taking any test image, and repeating the steps 1)-6) to obtain a segmentation prediction result The method is accurate in segmentation, low in calculation resource consumption and short in reasoning time.

Description

technical field [0001] The present invention relates to image processing, in particular to a fast high-resolution image segmentation method based on a block recommendation network. Background technique [0002] Since artificial intelligence is considered the fourth industrial revolution, the world's top and most influential technology companies such as Google and Facebook have turned their attention to AI. The contribution of neural networks in computer vision, natural language processing, etc. is unquestionable. With the continuous improvement of algorithms, research in some vertical fields has been applied. In the field of computer vision, the current applications of neural networks mainly include image recognition, target location and detection, and semantic segmentation. Image recognition is to tell you what the image is, target location and detection tell you where the target is in the image, and semantic segmentation is to answer the above two questions from the pixel...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to solve the problem that the semantic segmentation of high-resolution and ultra-high-resolution images is difficult to achieve a trade-off between computing resource consumption, reasoning speed, and accuracy, and to provide a method that can perform high-resolution and ultra-high-resolution images. Fast and accurate image segmentation with low memory consumption based on Patch Proposal Network (PPN) for high-resolution image segmentation

Images