Image display with uint8 and double

Can someone explain why the following image, 'yinyang.png', is displayed differently when I use uint8 and double in the imshow function?

    I = imread('yinyang.png');
    figure(1), imshow(I); title('image displayed with uint8')
    figure(2), imshow(double(I)); title('image displayed with double')

The original image displayed in Windows Photo Viewer and the two figure(1) and figure(2) are attached.

Answer 1

[Stephen23]

It seems that you have a "binary" image with values 0 and 1, which is stored as uint8.

• Because floating-point images are assumed to contain values between 0 and 1 the double image displays those colors as black and white.
• But for the uint8 image the values are assumed to be from 0 to 255, so those two colors are shown as black and very-dark-gray (indeed, the symbol is just visible in your screenshot).

You can probably use imshow's optional input argument to specify the (non-default) range:

imshow(double(I),[0,255])
imshow(I,[0,1])

See also:

https://www.mathworks.com/help/images/convert-image-data-between-classes.html

Addendum: the PNG standard supports 1-bit (i.e. binary) images, which MATLAB imports as logical images. Logical images are displayed as black-and-white.

[Mohsin Shah]

I checked it with imshow(double(I),[0,255]) and now both show the same display but why it is not shown the way it is shown in Windows photo viewer?

[Stephen23]

"why it is not shown the way it is shown in Windows photo viewer?"

We are not responsible for any image color processing, spatial anti-aliasing, or image scaling that might be performed by Windows Photo Viewer, and you should check its documentation. I certainly would not trust it as an accurate representation of an image. Do you?

From the screenshot that you provide it appears that the main difference comes from some spatial anti-aliasing: this means that Photo Viewer "smooths" the edges to make it look nicer for humans to look at... but this means what you see are not the "real" image pixel values. Most likely MATLAB is displaying the pixels "as they really are".

[Mohsin Shah]

Yes, I agree with you. I think Photo Viewer also do some sort of scaling for making the image appear more prominent and clear.

[Anindya Banerjee]

@ Stephen Cobeldick but inthe case of double, why the image turns out to be inverse? i.e white part of the image becomes black and black part becomes white?