Hi. I want increase the geometrical resolution of a thermal image by matching it with an RGB image of the same object, but keeping thermal information. How can I do?

Question

Federico Fioretti 2021-10-26

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1572248-hi-i-want-increase-the-geometrical-resolution-of-a-thermal-image-by-matching-it-with-an-rgb-image-o

编辑： Simar 2024-5-8

I have to create a 3D model of a building using thermal images and I need to increase geometrical resolution of the thermal images. I was thinking of doing it by matching the thermal images with corresponding RGB images captured by a digital camera. Doing this, i would keep the characteristics of thermal image, so temperature information, color palette, ecc. How can i do this?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Simar 2024-5-7

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1572248-hi-i-want-increase-the-geometrical-resolution-of-a-thermal-image-by-matching-it-with-an-rgb-image-o#answer_1453436

编辑：Simar 2024-5-8

在 MATLAB Online 中打开

Hi Federico,

I understand that you are creating 3D model of building using thermal images and aim to enhance geometrical resolution of the thermal images. This process is referred to as image fusion or multimodal image registration and fusion, which combines high-resolution spatial details from RGB images with thermal characteristics from thermal images.

Here is a step-by-step approach using MATLAB, which is well-suited for image processing and computer vision tasks:

Step 1: Image Acquisition

Ensure for every thermal image, a corresponding high-resolution RGB image should be available, ideally captured from the same or a similar viewpoint. This alignment simplifies registration process between two image types.

Step 2: Image Preprocessing

Preprocess both sets of images to make them more compatible for matching:

Geometric Alignment: If images originate from different viewpoints, geometric transformation is necessary for alignment. This alignment can be achieved either manually or via feature detection algorithms.
Rescaling: Adjust RGB images resolution to closely match that of thermal images prior to proceeding with additional processing steps.

Step 3: Image Registration

Image registration involves aligning two images- thermal and RGB, in this context ensuring corresponding points across images match.Image Processing Toolbox in MATLAB offers tools such as “imregister” for direct alignment or feature-based methods like “detectSURFFeatures” and “matchFeatures” for automated processing.

Example Code-

% Example of feature-based registration 
rgbImage = imread('rgbImage.jpg'); 
thermalImage = imread('thermalImage.jpg'); 
% Detect features 
points1 = detectSURFFeatures(rgbImage); 
points2 = detectSURFFeatures(thermalImage); 
% Extract features 
[features1, valid_points1] = extractFeatures(rgbImage, points1); 
[features2, valid_points2] = extractFeatures(thermalImage, points2); 
% Match features 
indexPairs = matchFeatures(features1, features2); 
matchedPoints1 = valid_points1(indexPairs(:, 1), :); 
matchedPoints2 = valid_points2(indexPairs(:, 2), :); 
% Estimate transform 
tform = estimateGeometricTransform(matchedPoints2, matchedPoints1, 'projective'); 
% Warp thermal image to match RGB image 
registeredThermalImage = imwarp(thermalImage, tform, 'OutputView', imref2d(size(rgbImage))); 

Step 4: Image Fusion

Following registration, subsequent phase involves fusing images. Objective is to improve resolution of thermal image while retaining its thermal characteristics. Techniques ranging from simple averaging to more advanced methods such as Laplacian Pyramids can be employed. These can be executed through fundamental operations.

Step 5: 3D Model Creation

Upon enhancing thermal images, construct the 3D model. integration of thermal information into 3D modelling process varies based on the intended application. Compatibility with chosen 3D modelling software may require exporting processed images in suitable format.

Step 6: Application of Thermal Images on 3D Model

Finally, map the enhanced thermal images onto the3D model. This could involve UV mapping or other techniques depending on the 3D software. Goal is to overlay thermal texture onto the model surfaces accurately.

This process requires image processing and some manual tweaking to get the best results. Please refer to the following documentation links-