How to fuse IR sensor Image with Visible Camera Image using MATLAB without Toolboxes?

Question

Zain ul abideen ahsan 2020-12-10

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/689824-how-to-fuse-ir-sensor-image-with-visible-camera-image-using-matlab-without-toolboxes

评论： Image Analyst 2020-12-12

So we have 2 images, One is IR Sensor image and the other is visible camera image with differnt size. I am not allowed to use any of Matlab toolboxes for fusing images

So what I seek is the output image of same size , filed of view and resolution as of IR image but with some features such as edges etc. added from visible image.

These are steps;

First Manipulate the visible image until it match the IR image (in size, filed of view and resolution)
Then extract features from manipulated visible image and combine it with features of IR image to output one single image with featurs of visible image but size and resoultion of IR image.

In my code i am resizing visible image manualy using imrize becuase I checked IR image was of 240x320 of size so i resized Visible image to this size manually.

I am attaching my code that i have written I am not sure if its perfect or not. Also attaching .mat file containg both IR image and Visible image saved

clc;
clear all;
 close all;
%Load images from .mat file
load('proj_IR.mat', 'IR')
load('proj_IR.mat', 'visible')
I1=IR
img= visible
%Resize visible image to size of IR image
I2 = imresize(img,[240 320]); 
imshow(IR)
tic
N=3;
Med1= medfilt2(I1, [N N]);
M=35;
h=1/(M*M)*ones(M);
b1=imfilter(double(I1),double(h),'circular');
d1=double(I1)-b1; 
S1=(b1-double(Med1)).^2;
Med2= medfilt2(I2, [N N]);
b2=imfilter(double(I2),double(h),'circular');
d2=double(I2)-b2;
S2=(b2-double(Med2)).^2;
w1=S1./(S1+S2);
w2=S2./(S1+S2);
F1=double(w1).*double(d1)+double(w2).*double(d2);
F2=0.5*b1+0.5*b2;
FF=double(F1)+F2;
toc
figure, imshow(I1, []);
figure, imshow(I2, []);
figure, imshow(S1, []);
figure, imshow(S2,[]);
figure, imshow(w1, []);
figure, imshow(w2, []);
figure, imshow(F1,[]);
figure, imshow(F2,[]);
figure, imshow(FF, []);
FF=uint8(FF);
figure,imshow(FF,[]);

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

Image Analyst 2020-12-10

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/689824-how-to-fuse-ir-sensor-image-with-visible-camera-image-using-matlab-without-toolboxes#answer_571454

编辑：Image Analyst 2020-12-10

You didn't resize the image or register it. What I'd try first is threshold each image to try to find the circles in each image. Then knowing the centers and diameter, paste (see attached demos) the smaller IR image onto a black canvass that is the same size as the visible light image. Then add the two together or put each into one color channel with imfuse().

4 个评论
显示 2更早的评论隐藏 2更早的评论

Image Analyst 2020-12-10

编辑：Image Analyst 2020-12-10

在 MATLAB Online 中打开

No. The method I'm thinking of involves only using < or > to do thresholding, then find() to find the circle coordinates, and mean() to find the center of the circle. Those don't require any toolboxes, just built-in functions.

mask = irImage < someValue;
[r, c] = find(mask);
xCenter = round(mean(c));
yCenter = round(mean(r));
xIRDiameter = max(c) - min(c);
yIRDiameter = max(r) - min(r);
row1 = round(xCenter - xIRDiameter/2);
col1 = round(xCenter - xIRDiameter/2);
row2 = row1 + yIRDiameter;
col2 = col1 + xIRDiameter;
subImage = irImage(row1:row2, col1:col2)

and so on for the visible light image. Then just paste the IR subimage onto a canvass of all zeros:

bigIR = zeros(size(visibleImage), 'uint8');

but you can't just paste it using the other coordinates, you have to adjust the coordinates to put the center where the center of the clock in the visible light image is. Plus you also have to resize it

subImage = imresize(subImage, [xVisibleDiameter, yVisibleDiameter]);

And when you paste that resized image you have to compute an upper left corner that is the same as the upper left corner for the clock block in the visible light image.

By the way you would not make measurements on an image that is a weighted sum of the IR and visible light image. It just doesn't make sense. I mean they don't even have the same units. The visible light image has units of gray levels while the IR image has units of temperature in degrees Celsius. What you could do is to use the visible light image to identify objects and produce a mask, then go to the IR image to measure the temperature within that mask.

Zain ul abideen ahsan 2020-12-10