how can i perform gray scale image normalization???

i want to implement normalization to gray scale image to reduce the effect of illumination's differences.

the eq. of the grayscale normalization is :

y=((x-min)*255/(max-min))

x : gray scale value of original image.

y : gray scale value of op image(after normalization).

min : minimum gray scale for the original image.

max : maximum gray scale for the original image.

i tried to perform this by :

m=imread();
min1=min(min(m));
max1=max(max(m));
y=((m-min1).*255)./(max1-min1);
imshow(m);figure,imshow(y);

but it is wrong code .

i dont know why ?

is there any help?

regards

[Image Analyst]

Perhaps because you didn't pass any filename into imread(). Perhaps because you left off the [] in imshow(y, []). It's floating so unless it's in the range 0-1 you need to give [] as the second arg to imshow().

I think you've been around here long enough to know that you need to tell us what the error message is (by copying and pasting the red text from the command window) because just saying "it is wrong code" is not sufficient for us to debug your code or figure out what is wrong or unexpected behavior.

[mmm ssss]

dear image analyst

what i means by it's wrong code that it did not enable me to get my goal .

the image y is only whole black image with only one white point inside it .

so this is not normalization

[Image Analyst]

Like I said, you probably just didn't display it correctly. See my answer below.

[Walter Roberson]

We would need to see the current code.

[Xylo]

you can use double() before main function....

y=double((m-min1).*255./(max1-min1)); and as m is a 2D variable, y should be 2D variable. i.e u have to write as for i=1:m for j=1:n y(i,j)=double((m(i,j)-min1).*255./(max1-min1)); end end

Answer 1

[Image Analyst]

Or you can simply do this:

normalizedImage = uint8(255*mat2gray(grayImage));
imshow(normalizedImage);

and not worry about the normalization because mat2gray will do it for you.

[mmm ssss]

not very good effect of mat2gray(grayimage)

[Image Analyst]

It does exactly what you asked for and exactly what all these other formulas here are doing. It maps the min of your array to 0 and the max of your array to 255. You can subtract the array from what you got with the other formulas and you'll see that everything is zero, meaning they're the same. If that doesn't have a "good effect" then none of the formulas here will either.

[mmm ssss]

image analyst

i think that your opininon is correct but, in many paper they use the grayscale normalization to reduce the differences in illumination.

[Image Analyst]

Then maybe their algorithm uses image normalization as just one step in the process and maybe you're not doing all the steps. Or else maybe their algorithm is not appropriate for the kind of video or images you have.

Answer 2

[Syed Ahson Ali Shah]

This is the Formula:

Normalized Image = (Original image - min of image) * ((newMax-newMin) / (ImageMax - ImageMin)) + newMin

where newMax and newMin is 255 and 0 respectively for the case when normalization is between 0 to 255.

[Image Analyst]

No it's not:

Originalimage = [100, 200]

Originalimage = 1×2

100 200

minofimage = min(Originalimage(:));
ImageMin = min(Originalimage(:));
ImageMax = max(Originalimage(:));
newMax = 255;
newMin = 0;
% Do the formula he gave.
NormalizedImage = (Originalimage - minofimage) * ((newMax-newMin) / (ImageMax - ImageMin)) + newMax

NormalizedImage = 1×2

255 510

% Do my formula:
normalizedImage = uint8(255*mat2gray(Originalimage))

normalizedImage = 1×2

0 255

[Syed Ahson Ali Shah]

There was typo mistake. I corrected now.

My answer is 100% correct. I guarantee

Answer 3

[Walter Roberson]

I would suggest you use

y = uint8(255 .* ((double(m)-min1)) ./ (max1-min1));

With your existing code, the (x-min) would be okay, but multiplying by 255 would get saturation to 255 whenever the difference was not 0, and then you would get integer division of that 0 or 255 by the range interval.

[mmm ssss]

??? Error using ==> minus

Integers can only be combined with integers of the same class, or scalar doubles.

i faced this matlab error

[Walter Roberson]

y = uint8(255 .* ((double(m)-double(min1))) ./ (max1-min1));

[mmm ssss]

y = uint8(255 .* ((double(m)-double(min1))) ./ (max1-min1));

??? Error using ==> rdivide

Integers can only be combined with integers of the same class, or scalar doubles.

[Walter Roberson]

You should be able to extrapolate.

y = uint8(255 .* ((double(m)-double(min1))) ./ double(max1-min1));

[mmm ssss]

can you see the modified code:

clear all

>> m=imread('E:\master_matlab\HandVein_DataSet\0010hv3.bmp');

min1=min(min(m));

max1=max(max(m));

y = uint8(255 .* ((double(m)-double(min1))) ./ double(max1-min1));

>> imshow(m);

>> figure,imshow(y);title(,normalization,);

i implemented also image analyst'method:

J = filter2(fspecial('sobel'), m);

K = mat2gray(m);

figure,imshow(K);

can you give me your opinion in the resultant images (Y&K) , is they good or the image before normalization is more suitable.