Unlock instant, AI-driven research and patent intelligence for your innovation.

Image patch matching method based on multi-scale convolution

A matching method and multi-scale technology, applied in the field of image processing, can solve the problem of not considering the multi-scale features of the patch, and achieve the effect of overcoming insufficient training, improving training efficiency, and eliminating differences.

Active Publication Date: 2020-01-07
XIDIAN UNIV
View PDF9 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

But this method does not consider the multi-scale features of the patch

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 patch matching method based on multi-scale convolution
  • Image patch matching method based on multi-scale convolution
  • Image patch matching method based on multi-scale convolution

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0084] S1. To evaluate the performance of our method, we validate it on the widely used homogeneous dataset UBC PhotoTour and the heterogeneous dataset VIS-NIR. UBC PhotoTour contains three subsets: Liberty, Yosemite, and Notredame. The three subsets contain 450K, 634K, and 468K independent patch blocks and 160K, 230K, and 147K unique 3D points. VIS-NIR contains 9 subsets, namely: Country, Field, Forest, Indoor, Mountain, Oldbuilding, Street, Urban, Water, and the matching samples and non-matching samples in each subset account for half. On the UBC PhotoTour dataset, we train on Liberty, Yosemite, and Notredame respectively, and then test on the other two subsets. On the VIS-NIR dataset, we train on the Country subset and test on the other 8 subsets;

[0085] S2. Each independent 3D point on the UBC PhotoTour dataset contains 2-5 patches, and 2 patches with the same 3D point form a matching pair. Randomly select 2 patches for each 3D point on each subset of UBC PhotoTour (T ...

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 patch matching method based on multi-scale convolution. The method comprises the following steps: preparing a data set; making the data set; preprocessing the data; designing a three-branch-double-channel network structure; designing a multi-scale convolution module; calculating similarity according to the extracted features; network training; predicting a matchingprobability; and evaluating network performance. According to the method, the problems of insufficient training, no use of multi-scale information and the like in the prior art are effectively solved, the performance of the network is greatly improved, the training efficiency of the network is improved, and the robustness of the network is enhanced. The method can be applied to the fields of image registration, image retrieval, image tracking, multi-view reconstruction and the like.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to an image patch matching method based on multi-scale convolution. Background technique [0002] Establishing an accurate matching correspondence between image patches plays a vital role in many computer vision fields, such as: image registration, image retrieval, fine-grained classification, etc. Since the appearance of the image is easily affected by many aspects such as viewing angle changes, illumination changes, occlusion, and camera parameter settings, image matching is very challenging, and the extracted features need to have good invariance and discrimination. In this paper we propose a general approach that not only achieves better matching results on homogeneous datasets, but also achieves state-of-the-art performance on more difficult heterogeneous datasets. [0003] Before deep learning, SIFT-based methods were widely used in the field of image mat...

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): G06K9/62G06K9/46
CPCG06V10/462G06V10/7557G06F18/214
Inventor 王爽焦李成魏少玮方帅杨博武李彦锋权豆
Owner XIDIAN UNIV