Using uint8 and Double

I have a input image of size 800*800 and of type uint8. When i do convolution with a 5*5 filter, it produces me a result of type Double. With the result of convolution, I have to perform some other operations and whether I must convert my convolution output to uint8 ? i have many variables which stores results of many operation. Whether all the variable should be of same type?

[John D'Errico]

So what is your question? You do what you need to do, for whatever problem you face at any given point. Sometimes that involves converting from an integer form to double or the reverse as necessary.