Salient target detection method based on super-large receptive field feature optimization

Abstract

The invention belongs to the field of computer vision, and provides a salient target detection method based on super-large receptive field feature optimization, which comprises the following steps: 1) taking ResNet-50 as a trunk frame to extract multi-scale feature information from an RGB image and then encoding; 2) optimizing the multi-scale features by using a super-large receptive field feature mechanism to generate high-quality features; and 3) performing complementarity fusion on the optimized features by using an invisible relation feature fusion mechanism, and then generating a final saliency map. Compared with the prior art, according to the salient detection method based on super-large receptive field feature optimization, super-large receptive field optimization multi-scale features are utilized, invisible relation feature fusion is carried out, and the multi-scale features are optimized layer by layer to generate a high-performance saliency map.

Description

Technical field: [0001] The invention relates to the field of image processing, in particular to a salient target detection method based on super large receptive field feature optimization. Background technique: [0002] The statements in this section only relate to the background technology related to the present invention, and do not necessarily constitute the prior art. [0003] With the rapid popularization of Internet technology, multimedia technology and digital products, digital images have become an important carrier for people to obtain information from the outside world. And this means that efficiently and quickly processing a large amount of digital image data has become a crucial issue. It is very inefficient for a computer to process all the details in a digital image simultaneously. If the limited computer resources are allocated to the salient target areas in the image, the efficiency of computer processing digital images will be greatly improved. Therefore...

AI Technical Summary

Problems solved by technology

However, the effect of further extracting semantic features by stacking convolutional layers, larger convolution kernels, or directly using dilated convolutions to increase the receptive field is not good.

Images