under flow and over flow of image

what is the pre-processing that executed on a gray image to prevent the over and under flow of image processing? the rang of image [0-255]

Answer 1

[Walter Roberson]

There is no method that can prevent all overflow and underflow in image processing. Even setting all image locations to zero is not enough, as "image processing" can include steps that add or subtract arbitrary constants from the image.

Answer 2

[Image Analyst]

You can cast your image to single or double if you don't want to have the intensities clipped to the range 0-255. That way if they are outside that range, they will still have the value they should have.

[Walter Roberson]

That will not be enough if, for example, one of the steps is to take exp() of the image values.

Given any particular image processing algorithm, it might be possible to do type conversion and scaling so as to avoid overflow and underflow, but there is no mechanism that can be used to avoid overflow and underflow for all image processing algorithms.

[Image Analyst]

Yes, there are limits. Personally, I've never run into a situation where I've ever exceeded the limits of double (where my values exceed realmax = 1.7977e+308), but I suppose one could.