Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Motion vector calculation method and device based on deep learning

A technology of motion vector and deep learning, which is applied in neural learning methods, computing, image data processing, etc., to achieve the effect of increasing accuracy, facilitating engineering realization, and increasing accuracy

Pending Publication Date: 2021-07-23
SHANGHAI TONGTU SEMICON TECH
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] In order to overcome the deficiencies of the above-mentioned prior art, the purpose of the present invention is to provide a motion vector calculation method and device based on deep learning to overcome the deficiencies of the traditional 3DRS technology in the calculation of small objects and solve the problem of small object image areas in video images. Various image quality issues in

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
  • Motion vector calculation method and device based on deep learning
  • Motion vector calculation method and device based on deep learning
  • Motion vector calculation method and device based on deep learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0089] Such as Figure 5 As shown, in this embodiment, a motion vector calculation method based on deep learning includes the following steps:

[0090] Step 1, training and identifying the tag information of the specified object in the image through the deep neural network, the tag information includes the size of the area where the object is located, the center coordinates, and the confidence level.

[0091] If the specified object is missed in the current frame, according to the inertial characteristics of the specified object, the position information of the object in the previous two frames is used to estimate the mark of the specified object in the current frame using a linear formula, and the estimated confidence information is based on the detection of the object in the previous two frames The confidence of the result is determined, and a certain penalty is added.

[0092] Step 2, knowing the label information of the specified object in the current frame and the previo...

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 a motion vector calculation method and device based on deep learning, and the method comprises the following steps: S1, recognizing the mark information of a specified object in a video image based on deep learning through employing a trained deep neural network; S2, estimating a motion vector MV of the specified object according to the identified mark information of the specified object in the front and back frame images; S3, expanding motion vector MV information obtained through estimation in the step (S2) by using an M * N window, and determining a motion vector MV'after refined processing according to an absolute difference value and SAD of each motion vector MV after expansion; and (S4) applying the motion vector MV 'after refined processing in the step (S3) to 3DRS so as to obtain a final interpolation motion vector MV.

Description

technical field [0001] The present invention relates to the technical fields of digital image processing, deep learning and artificial intelligence, in particular to a motion vector calculation method and device based on deep learning. Background technique [0002] MEMC (Motion Estimation Motion Compensation, motion prediction motion compensation) technology is an important interpolation technology in the field of FRC (Frame Rate Conversion, frame rate conversion). played a vital role. According to experimental research, the quality of the final image mainly depends on the correctness of MV (Motion Vector, motion vector) calculation, and partly depends on the quality of MC (Motion Compensation, motion compensation) interpolation. Considering the commercial value, MV calculation must not only ensure high accuracy, but also consider engineering implementation. The traditional 3-Dimension Recursive Search (3-Dimension Recursive Search, three-dimensional recursive search) metho...

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): G06T7/20G06N3/04G06N3/08
CPCG06T7/20G06N3/08G06T2207/10016G06T2207/20081G06T2207/20084G06N3/045
Inventor 王洪剑陈涛林江
Owner SHANGHAI TONGTU SEMICON TECH
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com