Video target segmentation method based on motion attention

Abstract

The invention provides a video target segmentation method based on motion attention, and the method comprises the steps: adding a channel feature map outputted by a channel attention module and a position feature map outputted by a motion attention module, and obtaining a segmentation result of a current frame, wherein the input of the channel attention module is the feature map Ft of the currentframe and an appearance feature map F0 of a target object provided by the first frame; enabling the channel attention module to calculate the association between the input feature map Ft and the F0 channel, wherein the output channel feature map reflects the object with the appearance closest to the target object in the current frame, the input of the motion attention module is the current frame feature map Ft and the position information Ht-1 of the target object predicted by the memory module in the previous frame motion attention network; enabling the motion attention module to calculate the association between the positions of the input feature map Ft and the Ht-1, wherein the output position feature map reflects the approximate position of the target object in the current frame. According to the invention, two factors of appearance and position are combined to realize more accurate segmentation of the video target.

Description

Technical field [0001] The invention belongs to the field of image processing and computer vision, and relates to a method for segmenting video objects, and in particular to a method for segmenting video objects based on motion attention. Background technique [0002] Video object segmentation is a prerequisite for solving many video tasks, and it plays an important role in object recognition, video compression and other fields. Video target segmentation can be defined as tracking the target and segmenting the target object according to the target mask. According to whether there is an initial mask, video target segmentation can be divided into semi-supervised and unsupervised methods. The semi-supervised segmentation method is to manually initialize the segmentation mask in the first frame of the video, and then track and segment the target object. The unsupervised method automatically segment the target object in a given video based on a certain mechanism without any prior inf...

AI Technical Summary

Problems solved by technology

[0005] The problem to be solved by the present invention is: in video object segmentation, only relying on the segmentation result of the previous frame to determine the target object in the current frame, the precise position of the target object cannot be obtained, and even due to excessive dependence on the segmentation result of the previous frame, resulting in The target object drifts; and the existing video target segmentation methods based on motion information, most of the target object segmentation is based on the optical flow information between the current frame and the previous frame, which not only has a large amount of calculation, but also limits the segmentation results specific sport mode

Images