Normalize image from -0.5 - 1.3 to 0 - 1

Hello,

I am processing an image through some filters and get the image with min value =-0.5014 and max=1.38. I would like to normalize it between 0-1 for comparative analysis using SSIM PSNR. The mat2gray function distorts the image. Please suggest how to work around this problem.

Thanks

Answer 1

[Ameer Hamza]

Try this

im_nrom = (im - min_val)./(max_val - min_val)

[fresta peng]

Thank you for the reply.

But it is the same as mat2gray.

When i do that i see a gray sheen on the image.

[Ameer Hamza]

What specific characteristics you want after normalization? Can you attach a sample image so that I can have a better idea.

[fresta peng]

please check the question . I added the images with corresponding values.

[Ameer Hamza]

fresta, If you directly used imread on image with [-0.5 1.3], the MATLAB directly truncate the value out of [0,1]. To get the same quality, try

im_new = im; % im is image in range [-0.5 1.3]
im(im<0) = 0;
im(im>1) = 1;

[fresta peng]

Yes the solution i guess here would be to clip the values like this. Thanks

[Image Analyst]

How is that a solution? Why do you want to clip your values? Is there any physical justification for doing that? Why not just use the original values and learn how to display things properly? How did you get values in that range anyway? Anyway, what things look like when they're displayed has no bearing on what image analysis you do. You could make the same image look like ANY brightness and contrast depending on what you put into imshow() between the brackets: imshow(grayImage, [lowValue, highValue]). It will map lowValue and darker of your image to pure black, and highValue and brighter to pure white. Values of your image in between lowValue and highValue will be linearly scaled in brightness between black (0) and white (255). You can do lots of things to make your image look better or pretty but that doesn't mean you need to, or should, change the underlying actual values. I recommend you don't change/clip the values unless there is some physical meaning/reason for doing so.

Answer 2

[Image Analyst]

mat2gray() does NOT distort the image spatially. I'd have to see proof of that (but I'm sure you won't be able to provide it). It just rescales the intensities. These codes will show the same image

subplot(1, 2, 1);
imshow(doubleImage, []);
subplot(1, 2, 2);
imshow(uint8Image, []);

even though the intensity ranges are different. In your imshow() you did not put a range in there, like [], so for the floating point image everything below 0 was clipped to 0 and everything above 1 was clipped to 1 (white) so it had more apparent contrast than the uint8 image.

Another function that doesn't require the Image Processing Toolbox is rescale

grayImage = rescale(grayImage, 0, 1);

[fresta peng]

Yes I think, like you said, the intensities are clipped. But these values are generated by filtering so the only solution is to simply round<0 values to 0 and >1 to 1.

Thank you all for the insights and help.