09-27-2011, 08:34 PM
[quote name='Brightcolours' timestamp='1317150136' post='11907']
But there is a big problem with downscaling. With downscaling you can not predict how the "noise" and the detail will behave/survive. Different sized images get downscaled different percentages, giving different results.
So, it is just wrong to do so for measurements.
If you want to test for same size prints, you will have to actually print them at the same size, without first down sampling the pixels to varying degrees. Else you are just "testing" down sampling algorithms....
[/quote]
Downsizing is just a linear factor so technically it is perfectly valid to do so for a proper comparison IMHO.
Eventually your normalization base gets a little outdated though but this will take a while.
But there is a big problem with downscaling. With downscaling you can not predict how the "noise" and the detail will behave/survive. Different sized images get downscaled different percentages, giving different results.
So, it is just wrong to do so for measurements.
If you want to test for same size prints, you will have to actually print them at the same size, without first down sampling the pixels to varying degrees. Else you are just "testing" down sampling algorithms....
[/quote]
Downsizing is just a linear factor so technically it is perfectly valid to do so for a proper comparison IMHO.
Eventually your normalization base gets a little outdated though but this will take a while.