Marking landmarks in image and determining real time distance panned by camera based on pixel value

I am using a camera to make recordings of the mouse cortex through the skull, but need to focus on certain areas. To find the area of interest, I need to know where I am relative to anatomical landmarks. I would be interested in seeing how far i travel from an easy to identify region anatomical region such as the bregma, to give me a more exact position of the location I am imaging.

Since the field of view of the camera is far smaller than the whole brain, I want to create a preview of my video, mark a spot in the brain in my preview as 0,0, and be able to pan the camera, for example 5 mm to the right, and keep track of how far the camera has panned relative to this spot by showing me real time updates of distance travleled.

This would mean that original spot would be out of frame and I need to somehow continously identify landmarks and keep track of how many pixels my image moved. Since I know how many pixels a mm corrosponds, the exact distance panned should be easy to calculate.

an example picture for context: https://imgur.com/a/GW3sHMJ

Answer 1

[Image Analyst]

This is what I see, and why we prefer that you attach the image here instead of on third party sites:

It's all black.

Perhaps you could use imregister() or imregcorr() or imregtform() to find out how much the camera shifted from one frame to the next.

[John Doe]

my apologies. here's the image once again uploaded to the website.

[Image Analyst]

Have you had a chance yet to try my suggestions? (Sorry, I don't have code for that.)

[John Doe]

I tried creating a script using imregister() , but it is very slow.

Here's an example of 34 frames extracted from a video panning over the cortex. each iteration takes around 1 minute, which is far too slow for any practical use.

I tried changing the optimizer settings, but with little success. Either its still too slow or I don't get an accurate measure of distance travelled.

x = 0
y = 0
figure
for i = 1:34
    %pic 1
    path = ('*');
    filename = ('*');
    prefix = strcat(path, filename);
    no = i;
    suffix = '.tif';
    filename = sprintf('%s%03i%s', prefix, no, suffix);
    
    pic_1 = imread(filename);
    
    
    
    
    
    
    
    %pic 2
    path = ('*');
    filename = ('*');
    prefix = strcat(path, filename);
    no = i+1;
    suffix = '.tif';
    filename = sprintf('%s%03i%s', prefix, no, suffix);
    
    
    pic_2 = imread(filename);
    
    
    [optimizer, metric] = imregconfig('monomodal');
    tform = imregtform(pic_1,pic_2,'similarity',optimizer,metric)
    
    
    optimizer.GradientMagnitudeTolerance = 1.000000e-04;
    optimizer.MinimumStepLength = 1.000000e-04;
    optimizer.MaximumStepLength = 6.250000e-02;
    optimizer.MaximumIterations = 50;
    optimizer.RelaxationFactor = 0.25;
    
    
    movingRegistered = imregister(pic_1, pic_2, 'affine', optimizer, metric); %perform registration
    
    
    
    
    tform.T;
    
    
    
    x = x-tform.T(3,1)
    y = y-tform.T(3,2)
    
    coordnates = strcat(' x: ',num2str(x),' y: ',num2str(y));
    RGB = insertText(pic_2,[20 20],coordnates,'FontSize',12,'BoxColor','green');
    
    
    imshow(RGB)
    
    
    
end

[John Doe]

I tried rescaling the picture to 1/10 its original size (1920x1200), to speed up the processing time using imresize(I,scale). While this did increase processing time significantly it still takes 10-20 seconds between each frame.

[Image Analyst]

You can hire the Mathworks to speed up your algorithm if it's really important to you. Otherwise you'll just have to live with it. I don't have the time to delve into it for you. Sorry.