the problem about use nlfilter

Hello, everyone

to imitate the demo from help and the demo from mathworks, I wrote the code about nlfilter.

    %a gray level image using nlfilter function
   A=imread('C:\Users\yanghang\Pictures\0.jpg');%  get the image dimension
   [rows,columns,numofColorBands]=size(A);
   if numofColorBands >1 %not the gray image
       grayimage=A(:,:,2) %Take Green channel
   end
   fun=@(x)median(x(:));% tell nlfilter to use function median to do its operation
   doubleImage=im2double(grayimage);
   B=nlfilter(grayimage,[3,3],fun); % filtering the image
   subplot(1,3,1);
   imshow(A);
   subplot(1,3,2);
   imshow(grayimage);
   subplot(1,3,3);
   imshow(B);

There is one thing troubles me. if I replace the first sentence with the second sentence

% doubleImage=im2double(grayimage);
doubleImage=double(grayimage);

I get two results. But in the workplace, the format is all double ,so why there are different?

Thanks in advance.

Hang Yang

Answer 1

[Walter Roberson]

() is approximately

function result = im2double(MATRIX)
   if isinteger(MATRIX)
     result = double(MATRIX) ./ intmax(class(MATRIX));
   else
     result = double(MATRIX);
   end
 end

(except a little more complicated for some more cases.)

That is, a uint8 image can have values 0, 1, 2, 3, ... 255. If you double() that then you would get 0.0, 1.0, 2.0, 3.0, ... 255.0 -- double precision numbers but large. However, images that are floating point need to be in the range 0.0 to 1.0, with 0.0 corresponding to minimum and 1.0 corresponding to maximum. To get from 0.0, 1.0, 2.0, 3.0, ... 255.0 to 0.0 to 1.0, you need to divide by 255.0, which is the maximum representable uint8 value.

[Y]

Thanks a lot