Rapid saliency detection method based on multi-scale feature attention mechanism

Abstract

The invention provides a rapid saliency detection method based on a multi-scale feature attention mechanism. The method comprises the following steps: firstly, processing an image through a deep convolutional network; obtaining features based on different convolution layers, including shallower layer features and advanced semantic features, then processing the obtained features based on differentconvolution layers, and finally inputting the processed shallower layer features and advanced semantic features into a decoder to generate a saliency detection graph. According to the invention, interference of most background features is eliminated; the calculation efficiency is improved; the background information is effectively inhibited; semantic information is better utilized for advanced semantic features by using pyramid expansion convolution, the features can be further refined by using a double-decoder junction, and the finally generated saliency map can completely highlight a saliency region in the image with a clear boundary and effectively suppress a background region.

Description

technical field [0001] The invention relates to the technical field of image and video processing, in particular to a fast saliency detection method based on a multi-scale feature attention mechanism. [0002] technical background [0003] With the popularization of wearable devices, smart phones and tablet computers with camera and video recording functions, the acquisition and storage of images and video information has become easier and easier, and people can shoot high-resolution images and video information of different durations at will. , so the number of images and videos has increased dramatically, which has also brought new challenges to research fields such as image and video processing. In recent years, studies have shown that the human visual system can quickly locate the most eye-catching objects from complex scenes, and how to use computer technology to simulate the human eye's visual mechanism and extract the human eye's interest area in images and videos has ...

AI Technical Summary

Problems solved by technology

To solve this problem, many object detection methods usually use feature representations representing image pyramids with multi-scale feature layers. However, the feature representations of image pyramids in current methods often integrate multi-scale convolution features indiscriminately. However, not all convolutions Product features are useful for saliency detection. Natural images contain complex foreground and background details, so low-level features often contain a lot of information. The redundancy of these feature details not only affects the calculation speed, but also affects The effect of the generated saliency map, so how to extract effective features and make reasonable use of them is a key point of saliency detection

Images