Combine multiple 2D images from a folder diretory into a 3D environment for visualisation.

Please help with fixing and extending my code, I would like to combine images of different size from different angles or views into a 3D visualisation. The 3D visualisation has to be real not only points.

% Load images from a folder 
imageFolder = 'Path';
images = imageDatastore(imageFolder); 
% Detect feature points in each image
imagePoints = cell(numel(images.Files), 1); 
for i = 1:numel(images.Files)
    I = readimage(images, i); 
    grayImage = rgb2gray(I); 
    imagePoints{i} = detectSURFFeatures(grayImage);
end
% Extract features and match them across images 
features = cell(numel(images.Files), 1); 
for i = 1:numel(images.Files) 
    I = readimage(images, i); 
    grayImage = rgb2gray(I); 
    [features{i}, validPoints{i}] = extractFeatures(grayImage, imagePoints{i});
end 
%Camera Intrinsics
focalLength = [10697 10770];
principalPoint = [2016 1512];
imageSize = [4042 3024];
intrinsics = cameraIntrinsics(focalLength,principalPoint,imageSize);
% Match features between consecutive images 
indexPairs = matchFeatures(features{1}, features{2});
matchedPoints1 = validPoints{1}(indexPairs(:, 1)); 
matchedPoints2 = validPoints{2}(indexPairs(:, 2)); 
% Estimate the fundamental matrix 
[F, inliers] = estimateFundamentalMatrix(matchedPoints1, matchedPoints2); 
% Compute the camera poses 
relativePose = relativeCameraPose(F, intrinsics, matchedPoints1(inliers), matchedPoints2(inliers)); 
% Reconstruct the 3D scene 
Points =[];
for i = 1:length(imagePoints)
    Points = vertcat(Points,imagePoints{i});
end
tracks = bundleAdjustment(Points , features, relativePose, intrinsics);
% Visualize the 3D model 
pcshow(tracks);
xlabel('X');
ylabel('Y'); 
zlabel('Z'); 
title('3D Model Reconstruction'); 

[Image Analyst]

I don't know what your code does but aren't you essentially describing computed tomography (CT)? For that you use filtered back projection. With projections from enough angles you can reconstruct the 3-D object. I'm not familiar with the functions you use so I don't know if they can do a CT-like filtered back projection.

https://howradiologyworks.com/filtered-backprojection-fbp-illustrated-guide-for-radiologic-technologists/

https://www.google.com/search?client=firefox-b-1-d&q=filtered+back+projection+ct+scan

The math for it won the inventors the Nobel Prize: https://www.nobelprize.org/prizes/medicine/1979/summary/#:~:text=Hounsfield-,Prize%20share:%201/2,development%20of%20computer%20assisted%20tomography%22

[Chumani]

Good day,

thank you for your response, i was not aware about this recent CT study, i will check. however, my is suppose to extract features of the image which will be SURFPoints in a single data type and then use camera calibration (which i used examples in this case for an iphone 13 pro camera at i used it to take pictures) then calculate camera poses infact view angles and then recontruct the image using bundleadjustment function to 3D points. Then i want to plot those 3D points to visualize a 3D image that i can rotate 360 degrees. From then i plan to extract target image points, the goal is to combine all 2D views images of same object and make a 3D of an object.

I am using vSLAM — Functions:

https://www.mathworks.com/help/vision/referencelist.html?type=function&s_tid=CRUX_topnav&category=visual-simultaneous-localization-and-mapping-slam

[Chumani]

the cat function wont do what i want as it stacks image slices and i take images from different views of either z, y,x planes not only z plane