If all the images have the same geometry, you can do it directly:
inpict=imread('explot.jpg'); % i renamed the image
croppedpict=inpict(39:294,169:451,:);
You could just do it in a loop as required.
If the image geometries vary and you need to programmatically crop the desired area, the process would differ. I'm sure there are vastly better ways of doing it, but since I feel lazy, I'm going to resort to just using the tools I'm familiar with most.
inpict=imread('explot.jpg'); % i renamed the image
% creat a thresholded copy
mpict=rgb2gray(inpict)<0.3*255;
% get rid of all objects except the largest one
mpict=bwareafilt(mpict,1);
% get the location of the ROI
[~,width]=cropborder(mpict,[NaN NaN NaN NaN]);
% use that location to extract the ROI
croppedpict=cropborder(inpict,width);
imshow2('croppedpict','tools')
cropborder and imshow2 are from the MIMT on the File Exchange:
If you don't have bwareafilt() because you don't have IPT, you can use bwareafiltFB from MIMT.
EDIT: I guess I can add a non-mimt way of doing it...
% generate a simple mask to isolate the ROI
mpict=rgb2gray(inpict)<0.3*255;
mpict=bwareafilt(mpict,1);
% find the location of the object edges
rows=max(mpict,[],2);
cols=max(mpict,[],1);
y=[find(rows,1,'first') find(rows,1,'last')];
x=[find(cols,1,'first') find(cols,1,'last')];
% use that location to extract the ROI
croppedpict=inpict(y(1):y(2),x(1):x(2),:);
It works. Still uses IPT though.
