国際・国内特許データベース検索

1. (WO2018176926) REAL-TIME CORRECTION METHOD AND SYSTEM FOR SELF-LEARNING MULTI-CHANNEL IMAGE FUSION

Pub. No.:    WO/2018/176926    International Application No.:    PCT/CN2017/117123
Publication Date: Fri Oct 05 01:59:59 CEST 2018 International Filing Date: Wed Dec 20 00:59:59 CET 2017
IPC: G06T 5/50
G06T 5/00
Applicants: SHANGHAI XUNMO COMMUNICATION TECHNOLOGY CO., LTD.
上海讯陌通讯技术有限公司
Inventors: ZHAO, Fengping
赵凤萍
DENG, Jinfu
邓金富
Title: REAL-TIME CORRECTION METHOD AND SYSTEM FOR SELF-LEARNING MULTI-CHANNEL IMAGE FUSION
Abstract:
A real-time correction method and system for self-learning multi-channel image fusion is provided, comprising: first, extracting and matching a contour and an object feature in the field of view, learning to generate a matching template applicable to fast fusion, and establishing a fusion matrix of an initial evaluation block according to a pre-segmented macroblock of an image; following a fixed frame rate to obtain, at one time, multi-channel image data of multiple points in time, and combining the data with the matching template; evaluating the combined image, and determining whether the fusion is successful; and if so, reducing a minimum macroblock of an original image feature area by one level to generate a template for further improvement of accuracy and recognition. According to the commonality of feature matrices of multiple adjacent frames and by determining the change trend of the feature matrices, different known feature matrices and correction methods are used, so as to rapidly combine and splice images from multiple live video streams, realizing real-time processing of video images, and stably and continuously splicing multiple real-time video image streams.