How to subtract a 16 bit image from 8 bit image?

How to subtract a 16 bit image from 8 bit image?

[Geoff Hayes]

kumar - please provide more details on what you are trying to accomplish. Why would you want to subtract a 16-bit image from an 8-bit one? What problem are you trying to solve, and have you written (or made an attempt at writing) any code?

[Image Analyst]

What is the range of values that the 16-bit image takes on? Are they in the 0-255 range, or are they greater than that?

[kumar Saurav]

I am attaching two images. One of them is 8 bit and other one is 16 bit. Background_subtract image is an 8bit image which I get as background from an infrared camera and after my experiment I get another image named L3_cavity_emission. I would like to remove the background but I couldn't succeed as I saved both image in different bits. Can you help me with it please?

[kumar Saurav]

16-bit image range is from 0 to 2^16-1. 0-255 is range for 8 bit image.

Answer 1

[Image Analyst]

Then, if the 16 bit image takes up the whole range, just read in the 16 bit image and divide by 256. Then subtract the 8 bit image from that, or vice versa. Cast to double if you want to get both positive and negative numbers.

[kumar Saurav]

I did this but then maximum of converted image if not 255 anymore. It's close to 60. Here is the code that I am using

a = imread('background_subtract.png');

b = imread('L3_cavity_emmison.png');

b8 = round(double(b)/255);

a8=a(:,:,1);

newimage=uint8(b8-double(a8));

image(newimage)

colormap 'gray'

[Image Analyst]

To scale each image to between 0 and 255, try this:

a8 = uint8(255*mat2gray(a));
b8 = uint8(255*mat2gray(b));
differenceImage = b8-a8;
imshow(differenceImage, []);
colormap(gray(256));

Answer 2

[Adam]

a = imread('background_subtract.png');
b = imread('L3_cavity_emmison.png');

a = double( a(:,:,1) ) * 255;

diff = double(b) - a;

That seems to remove the background as expected when I tried it just now. Going the other way down to 8-bit size should work equally well so long as you cast to doubles in the right place and don't do maths as integer before the cast.