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

Video high frame rate reproduction method based on grid structure deep learning

A technology of deep learning and grid structure, applied in the direction of neural learning methods, neural architecture, interpolation processing conversion, etc., can solve problems such as fuzzy synthesis results, unsatisfactory performance performance, and motion blur

Active Publication Date: 2018-11-16
福建帝视信息科技有限公司
View PDF5 Cites 11 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

High frame rate reproduction algorithm based on optical flow estimation For video scenes with motion blur and fast motion, it is difficult to estimate a very accurate optical flow
In addition, the performance of the spatial adaptive convolution method for video scenes with occluders is not satisfactory, and the synthesis results are usually blurred and disordered

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
  • Video high frame rate reproduction method based on grid structure deep learning
  • Video high frame rate reproduction method based on grid structure deep learning
  • Video high frame rate reproduction method based on grid structure deep learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0058] Such as Figure 1-4 As shown in one of them, the present invention discloses a method for reproducing video at a high frame rate based on grid structure deep learning, which is divided into the following steps:

[0059] Step 0, image selection for training database. The training data set of this patent is the UCF-101 action data set [5] , which covers more than 10,000 action videos. We randomly sample the video, and select high-quality video frames with obvious motion (the selection criterion of the present invention is to consider PSNR greater than 35 as high-quality images). Finally, 24,000 sets of video frames were selected, and each set consisted of three consecutive images.

[0060] Step 1, the production of the training database, resets the image size of the selected training data. First set the original image uniformly to the size of H*W, then normalize the image to the [-1,1] interval, and finally form a paired set containing N images where c∈{1,2,…,N}, H ...

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 video high frame rate reproduction method based on grid structure deep learning. A three-dimensional pixel stream estimated by adopting a grid structure mode can obtain a more accurate result in motion scenes of various motion quantities. Compared with the prior art, the method of the invention is more robust. In order to further improve the precision of the three-dimensional pixel stream and the effect of high frame rate reproduction, the invention proposes a way of combining a convolution feature extraction layer and a grid network structure. Compared with other prior art, the result of the high frame rate reproduction obtained by the method of the invention is more delicate and realistic in the detail texture of a synthesized frame.

Description

technical field [0001] The invention relates to the field of video high frame rate reproduction, in particular to a method for video high frame rate reproduction based on grid structure deep learning. Background technique [0002] Video high frame rate reproduction is to use the video image information of adjacent frames in the video sequence to estimate the key frame in the middle, which belongs to a classic image processing problem. In general, video high frame rate reproduction algorithms can be divided into interpolated frames and extrapolated frames. The former is to use the information of two consecutive frames of images to estimate the key frame in the middle; the latter is to use the information of two consecutive frames of video images in the video sequence to estimate the previous frame or the next frame. [0003] According to the continuous video image information in the video sequence, the video high frame rate reproduction algorithm is a method to reasonably us...

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
IPC IPC(8): G06T5/00G06T7/207G06T17/20G06N3/04G06N3/08H04N7/01
CPCH04N7/0135G06N3/084G06T7/207G06T17/20G06N3/045G06T5/73
Inventor 刘文哲李根童同高钦泉
Owner 福建帝视信息科技有限公司
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