How to do background subtraction between two images to find the ROI?

I took two frames,I want to subtract the background so that I can work on the region which I am Interested(Human). I have attached two pictures.

is the image of a road.The image with a human is this

Please help me with the code.I want to separate the human and crop the human region.

Thank you

Answer 1

[Image Analyst]

Just subtract

diffImage = double(rgbImage1) - double(rgbImage1);

and threshold above the noise level (say it's 10 gray levels.

mask = abs(diffImage) > 10;

To get a mask where any color channel is more than 10:

mask = any(mask, 3);

To get a mask where all color channels are more than 10:

mask = all(mask, 3);

Now mask is a 2-D logical array.

[Jawad Siddique]

if true
  clc
clear all;

I=imread('C:\Users\ASUS\Desktop\Project\pdanik.jpg');

J=imread('C:\Users\ASUS\Desktop\Project\road1.jpg');

M =imresize(I,[512 256]);

N =imresize(J,[512 256]);

I1 = im2double(rgb2gray(M));

I2 = im2double(rgb2gray(N));

diffImage = I1 - I2;

mask = abs(diffImage) > 10;

mask = any(mask, 3);

mask = all(mask, 3);

imshow(diffImage);

end
The output is not as expected

[Image Analyst]

See attached m-file.

Adapt as needed. For example you might call imregister to align the two images better, or call bwareafilt() to get rid of blobs smaller than a certain size.

[Jawad Siddique]

Thanks a lot,you are the best.I detected my ROI.Now i want to xtract the ROI for further processing.Anybody to help me out.

if true
  function testRGBImage()
clc;    % Clear the command window.
close all;  % Close all figures (except those of imtool.)
rgbImage1 = imread('C:\Users\ASUS\Desktop\Project\road.png');
rgbImage1 = rgbImage1(657:1263, :, :); % Crop black off top and bottom.

[rows, columns, numberOfColorChannels] = size(rgbImage1);

subplot(2, 2, 1); imshow(rgbImage1, []);

rgbImage2 = imread('C:\Users\ASUS\Desktop\Project\Human.png'); rgbImage2 = rgbImage2(657:1263, :, :); % Crop black off top and bottom.

[rows, columns, numberOfColorChannels] = size(rgbImage2);

subplot(2, 2, 2); imshow(rgbImage2, []);

diffImage = abs(double(rgbImage1) - double(rgbImage2));

thresholdValue = 60; mask = diffImage > thresholdValue;

mask = any(mask, 3);

 mask = all(mask, 3);

subplot(2, 2, 3); imshow(mask, []);

rp = regionprops(mask, 'BoundingBox', 'Area'); area = [rp.Area].'; [~,ind] = max(area); bb = rp(ind).BoundingBox; subplot(2, 2, 4); imshow(rgbImage2); rectangle('Position', bb, 'EdgeColor', 'red');

end

[Image Analyst]

If you want to extract just the largest blob, use

mask = bwareafilt(mask, 1);

But I can't write a whole turnkey app for you. It would still take a lot of work from here. Why don't you just use the built-in code for finding things: https://www.mathworks.com/products/computer-vision/features.html#object-tracking-and-motion-estimation The computer Vision System Toolbox does what you want to do, I think.

[Jawad Siddique]

I did that and got the largest blob,I detected the region in the original image where there is the largest blob,I want to extract the region bounded on the rectangle to another subimage for further processing.Badly in need of some help.The following code extract some region but not all.

if true
  subImage = rgbImage2(round(rp(ind).BoundingBox(2):rp(ind).BoundingBox(2)+rp(ind).BoundingBox(4)),round(rp(ind).BoundingBox(1):rp(ind).BoundingBox(1)+rp(ind).BoundingBox(3)));
end

Thank you.

[Image Analyst]

It does not need to be cropped for further processing. You can do everything you need to do on the original image. If you want to get the head, you'll have to tweak some other parameters, like threshold, or use additional operations like imclose() or others.

[Jawad Siddique]

Is there any way to crop out that region?It will really help me to get to the next stage. Thank you.

[Tanuja Minde]

can anyone send the whole code plz ? for the above operation.

[Image Analyst]

Do you mean like this (I'm just copying and pasting):

% Just subtract
diffImage = double(rgbImage1) - double(rgbImage1);
% and threshold above the noise level (say it's 10 gray levels.
mask = abs(diffImage) > 10;
% To get a mask where any color channel is more than 10:
mask = any(mask, 3);
% To get a mask where all color channels are more than 10:
mask = all(mask, 3);
% Now mask is a 2-D logical array.

Obvously you'll have to call imread() or whatever to read in your own rgbImage1 and rgbImage2.