aligning overlapping images and averaging the overlap

5 views (last 30 days)
i have an empty matrix
canvas = uint8(zeros(10));
images are placed inside
image1=uint8(magic(3));
canvas(2:4,2:4)=image1;
a second image obviously overwrites the first
image2=uint8(magic(3));
canvas(3:5,3:5)=image2;
I would like to average the overlapping region
my attempt was to inspect each element in canvas to see if it was zero or non zero and then average each element(of subsequent images) one by one if it was non zero. sadly my code is really slow and doesn't work properly..
please help

Accepted Answer

David Young
David Young on 10 Mar 2011
I'm not quite sure what you want to do in the non-overlapping region, but if you want to leave the original images as they are, then you can do it like this:
canvas1 = uint8(zeros(10));
canvas2 = canvas1;
image1=uint8(magic(3));
canvas1(2:4,2:4)=image1;
image2=uint8(magic(3));
canvas2(3:5,3:5)=image2;
canvas = canvas1 + canvas2;
overlap = canvas1 & canvas2;
canvas(overlap) = canvas(overlap)/2;
I don't think this corresponds exactly to your code though. If you could say what the result in your example should be, it would be possible to be more certain of what to do.
  2 Comments
james
james on 10 Mar 2011
this does what i want, it should be much faster than the for loops.
thank you
David Young
David Young on 10 Mar 2011
Just a thought - using uint8 for all the arithmetic might cause you problems. For example, it will mean that no value in the overlapping region can exceed 128. It might be wise to do all the arithmetic in floating point, if possible, and convert back to uint8 when definitely need the image in that format.

Sign in to comment.

More Answers (3)

james
james on 10 Mar 2011
this method doesn't work..
canvas = uint8(zeros(10));
image1=uint8(magic(3));
image2=uint8(magic(3));
canvas(2:4,2:4)=image1;
for rowcounter=1:3;
for colcounter=1:3;
if canvas(rowcounter,colcounter)==0;
canvas(2+rowcounter,2+colcounter)=image2(rowcounter,colcounter);
elseif canvas(rowcounter,colcounter)~=0;
canvas(2+rowcounter,2+colcounter)=((canvas(2+rowcounter,2+colcounter))+(image2(rowcounter,colcounter)))/2;
end
end
end
image(canvas)

Sean de Wolski
Sean de Wolski on 10 Mar 2011
To start: -use CONV2 with a ones kernel to figure out how many zeros are touching.
  2 Comments
james
james on 10 Mar 2011
can you put that in the context of the ^^ code?
Sean de Wolski
Sean de Wolski on 10 Mar 2011
No, since your above code "doesn't work" for you. I would guess what is wrong with your above code is that you're updating things and then checking the updates but I don't really know. CONV will do the averaging for you; you can use logical indexing to find if an element is zero. I don't totally understand what you're trying to do though, so more explanation would be better.

Sign in to comment.


james
james on 10 Mar 2011
this works, could it be faster?
canvas = uint8(zeros(10));
image1=uint8(magic(3));
image2=uint8(magic(3));
canvas(2:4,2:4)=image1;
for rowcounter=1:3;
for colcounter=1:3;
if canvas(2+rowcounter,2+colcounter)==0;
canvas(2+rowcounter,2+colcounter)=image2(rowcounter,colcounter);
end
if canvas(2+rowcounter,2+colcounter)~=0;
canvas(2+rowcounter,2+colcounter)=((canvas(2+rowcounter,2+colcounter))+(image2(rowcounter,colcounter)))/2;
end
end
end
image(canvas);
  2 Comments
Sean de Wolski
Sean de Wolski on 10 Mar 2011
You do realize that the it will add image2 twice since you're using two if statements not an else-if statement right?
james
james on 10 Mar 2011
never assume! no i didn't realise thanks.

Sign in to comment.

Categories

Find more on Deep Learning Toolbox in Help Center and File Exchange

Products

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!