# DP2M vs. 5DMkII: The DP2M realy is that good ...(large image)

Started Oct 27, 2012 | Discussions thread
 Like?
 Re: DP2M vs. 5DMkII: The DP2M realy is that good ...(large image) In reply to glacierpete, Oct 28, 2012
because the current upsampling interpolation algorithms are for bayer images.

Most resampling methods are based on the attempt to create an optimal low pass filter that windows the sinc to reduce ringing. The optimality of the sinc filter (which is the Fourier transform of the box function) can be easily (well it was easy for me during the signal theory lecture but that was 10 years ago) proven. This is the only(!) filter that exactly reconstructs a signal that is limited in frequency to half the sampling rate. (It is a bit hard for me to explain because I am not a native speaker and miss the right subject-specific vocabulary.) Unfortunately the sinc filter has an infinite carrier (is that the right term? im German it is TrĂ¤ger) and thus both inefficient to apply and introduces ringing around edges. So that's why a window function is applied... the most popular one results in the well-known Lanczos resampling which is a high quality resampling function usually used in image resizing.

Have a look at:

http://en.wikipedia.org/wiki/Fourier_transform

http://en.wikipedia.org/wiki/Sinc_filter

http://en.wikipedia.org/wiki/Window_function

http://en.wikipedia.org/wiki/Lanczos_resampling

And that is pure math - a ground truth that holds regardless of Foveon and Bayer and whatever image sensor.

So where does you claim stem from?

(Specific upsizing method like the one you mentioned or fractal upsizing applies some statistical model on the image like edge detection, pattern recognition on top of a sinc * window function.)

Joerg V's gear list:Joerg V's gear list
Sigma DP1 Nikon D800 Nikon AF-S Micro-Nikkor 60mm f/2.8G ED Nikon AF-S Nikkor 50mm f/1.8G Sigma 35mm F1.4 DG HSM +3 more
Complain
Post ()
Keyboard shortcuts: