Can someone help me understand this code? It's a crop function for image processing.

I'm doing a project on gesture recognition. A part of the code is this function imcrop, as given below. Can someone explain how this works? And also how I can integrate it into the main code?

% MATLAB FUCTION imcrop TO CROP THE CAPTURED IMAGE

function imgout=imgcrop(imgin) %function definition

imgin = imresize(cropimg,[240,240]); %resize the original image

columnsum=sum(cropimg); %find the sum of column entries

rowsum=sum(cropimg'); %find the sum of row entries

q1=1;

q2=240;

q3=1;

q4=240;

for w=1:240 %search for the vertical cropping boundary

if (columnsum(1,w)>=10)

q1=w;

break;

end

end

for w=q1:240 %search for the vertical cropping boundary

if (columnsum(1,w)>=10)

q2=w;

end

end

for w=1:240 %search for the horizontal cropping boundary

if (rowsum(1,w)>=10) q3=w;

break;

end

end

for w=q3:240 %search for the horizontal cropping boundary

if (rowsum(1,w)>=50)

q4=w;

end

end

%crop the image between boundaries and resize to original

imgout=cropimg(q3:q4,q1:q2);

imgout=[zeros(q4-q3+1,160-(q2-q1+1)),imgout;zeros(120-(q4-q3+1),160)];

imgout=imresize(imgout,[120,160]);

imgout=double(imgout); %convert the image type to double for future use

Answer 1

[Image Analyst]

It doesn't work. As soon as it hits the first line

imgin = imresize(cropimg,[240,240]);

it will throw an exception because cropimg is undefined. It looks like something written by a very novice MATLAB programmer to find the rows and columns where the image is bright enough to be retained. Instead of all those for loops, they could simply use find(). Then they pad the image with zeros on the left and right and resize it for some reason. Basically they're masking out dark stuff outside the bounding box but doing it in a very odd way. If you want a masking app, just ask me for a demo.

[Anjana Rao]

That would be so helpful! Thank you so much!

The project involves using Neural Networks for pattern recognition. The input and target images are cropped and resized to 240x240 to make the training happen faster. Maybe that's why the image is resized?

[Image Analyst]

See attached masking demos.

[Anjana Rao]

Thank you! :)