Question: Why do we need to sit under a piano?
Answer: Most of the time we don't need to do that. But there might be times it is useful to sit under a piano. It would depend upon what you wanted to do. I doubt anyone could come up with a complete list of reasons why anyone would sit under a piano.
Normalizing data is the process of taking data that might exist over a wide range, and doing a calculation on the data that results in the transformed data being contained within a pre-defined interval that happens to be useful for calculating with.
Some of the common normalization procedures include:
- subtract the mean from the data, so that the transformed data has a mean of 0
- divide the data by its standard deviation, so that the transformed data has a standard deviation of 1
- subtract the minimum of the data, and then divide that result by the new maximum, so that the transformed data covers the exact interval 0 to 1 inclusive
If you had a calculation which works on data with a mean of 0 and a standard deviation of 1, then it is common that you can usefully apply the algorithm to any data set by subtracting the mean and dividing by the standard deviation and working with the result.
I am going to go out on a limb and guess that you are working with Feature Detection for something. If that is the case, then the reason you normalize is so that you can deal with images of different sizes, so that (for example) your code is able to detect a circle with radius 51 just as easy as it can detect a circle with radius 4.
