Image foreground segmentation method combining subject determination and edge precision

A foreground segmentation and combination technology, applied in the information field, can solve the problems that segmentation results are prone to over-segmentation, universality and speed need to be strengthened, etc., and achieve the effect of improving detection accuracy and accuracy.

Pending Publication Date: 2021-11-02
ZHEJIANG UNIV OF TECH
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, the segmentation method based on low-level feature information has great limitations: the image needs to have similarity and continuity in the same area, the quality of the segmentation effect is strongly dependent on the similarity condition, and the segmentation result is prone to over-segmentation
However, the development of segmentation methods based on high-level feature information is still immature, and its universality and speed still need to be strengthened

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Image foreground segmentation method combining subject determination and edge precision
  • Image foreground segmentation method combining subject determination and edge precision
  • Image foreground segmentation method combining subject determination and edge precision

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0067] A method of determining the foreground image segmentation method and precision of the edges of the body in combination, comprising a main body determining process, a combination of both processes and process precision of an edge.

[0068] 1) determining the subject as follows:

[0069] 1.1) the input image using context-aware pyramid feature extraction module (the CPFE) for high-level multi-scale characteristic map, in order to obtain a rich context feature. (Which CPFE model for an existing model, are presented in CVPR2019 Conference). CPFE the VGG-16 network architecture Conv3-3, Conv4-3 and Conv5-3 as the basic high-level features, which VGG-16 is developed with Computer Vision Group at Oxford University researcher and Google's 16 layers deep convolution nerve network, VGG-16 comprises 13 layers convolution (convolutional five blocks, each block comprising convolution convolution layers 2-3) and three fully connected layers. Conv3-3 a third convolutional blocks inside th...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses an image foreground segmentation method combining main body determination and edge precision, which combines a main body determination process for positioning a salient region and an edge precision process for precisely segmenting a target. Firstly, a main body part is determined, and a context sensing pyramid feature extraction module is designed to obtain rich context features; a channel attention mechanism CA module after feature mapping of a context sensing pyramid feature extraction module and a space attention mechanism SA module after low-level feature mapping are combined, and cross entropy loss is used for supervising generation of significant boundary positioning information so as to obtain positioning information; secondly, an edge process is precise, non-local color features in the image are obtained on the basis of a spectrum extinction technology, advanced semantic features are obtained through a ResNet-101 deep residual network, pixel points in the image are classified by combining the non-local color features with the advanced semantic features through a Laplacian matrix, and the effect of precise target segmentation is achieved; and finally, the results of the two processes are fused.

Description

Technical field [0001] The present invention relates to the field of information technology, and more particularly to a method of determining an image foreground segmentation method combining the body and precision edge. Background technique [0002] When people get an image, the more introduced partial part is often a significant main area, which is called the foreground. Some spatial information (color, contour, texture, grayscale, etc.) in the foreground part is different from the surrounding environment. The foreground segmentation is based on these differences, and the image will split and extract the image with unique properties. Such as FASR R-CNN, Mask R-CNN, Masklab has a wide range of applications in people's actual life: the measurement image in the medical field, in the field of cloud map in the field of remote sensing, in the transportation field Extraction of vehicle contour characteristics and pedestrian testing, etc. While foreground segmentation technology has fa...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06T7/194G06T7/12G06T7/90G06K9/46G06K9/62G06N3/04G06N3/08
CPCG06T7/194G06T7/12G06T7/90G06N3/08G06T2207/20016G06T2207/20081G06N3/045G06N3/048G06F18/23213G06F18/253
Inventor 陆佳炜朱冰倩陈纬鉴姜钦凯董振兴朱明杰程振波
Owner ZHEJIANG UNIV OF TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products