why we need to convert the uint8 to double

double()

[Stephen23]

We only "need to convert the uint8 to double" if an operation requires this. You have not told us what operations you are doing, so we cannot tell you why you "need" to convert to double.

Answer 1

[Star Strider]

Most images are uint8, and this limits most operations to 8-bit precision, if they can be done at all. (Many functions accept only double-precision arguments.) You need to convert them to double to perform certain calculations on them such as fft2, fftn, or other such operations.

[Walter Roberson]

In particular if you want to do image subtraction then you need to convert to a data type that supports signed values.

Also if you want to calculate ratios then you need a floating point data type as ratios of integers immediately rounds to 0 or 1.

[Image Analyst]

conv2() is also a really common image processing routine that requires doubles. Summing also would if you would exceed the max value allowable for that integer type.