Am I oversharpening
in [Digital Cameras]
|
From: mark on 18 Jul 2005 09:45 JPS(a)no.komm schreef: > In message <1120808817.538631.149690(a)o13g2000cwo.googlegroups.com>, > "mark" <mspan(a)mad.scientist.com> wrote: > > >I'm not going there (with you at least). > > Why not? > > >Bottom line is that when you > >take raw information from the sensor, bayer sensors give you 1/3 color > >info on each site, > > Actually. there is no loss of color coverage at all; the AA filter makes > sure that a high percentage of light of all colors hitting anywhere on a > particular point of the sensor plane reaches the three color sensors. So, in your opinion, on a bayer sensor the light that falls on a single (say: green) input pixel also falls on its nabouring (red, blue) input pixels? So in practice, a group of input pixels records the same 'ray of light', allthough thay are in fact located on seperate positions on the sensor. This is what the AA filter does? > It is just not possible to record color details above a certain > frequency, which is a bit lower than the frequency of luminance detail > recorded. We generally can't see color detail at those frequencies > anyway, with normal viewing situations. And this luminance detail is recorded because the fraction of light that does not belong to the 'high percentage of light of all colors' is captured by the input pixels individually? > > >foveon gives you all. Debayer software interpolates > >the color info to create those humongous large files. > > Foveon does not give all. Foveon gives partial. Foveon has three color > channels, but they are not particularly distinct, and the hue resolution > can be pretty coarse, especially in the blue/green range. And the > blue/green seems to often vary in large, obvious blotches, rather than > in a manageable fine dither. > > Also, none of the current x3f-based cameras have proper aliasing, so > they do not give you full spatial information. On the SD9, the capture > has a very strong "snap-to-grid" effect, at the pixel level, and this > makes the images look artificial to me. There are lots of artifacts > that alighn themselves in horizontal and vertical lines, something that > does not happen in real vision. > As real vision is analogue? > If the Sigma cameras had proper AA filters, they would be better, IMO, > but a lot of the people drawn to them might not like them, because they > like the artificial sharpness of aliasing. > > You can reproduce the effect easily with any Bayer camera; upsample with > bicubic to twice the dimensions of a Sigma image (4536*3024), and then > downsize to 2268*1512 with Nearest Neighbor. Voila! Instant SD9 "pop" > and staircase effects. tried that. no way. Honest questions! > -- > > <>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> > John P Sheehy <JPS(a)no.komm> > ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
From: mark on 21 Jul 2005 03:58 John, where are you? mark schreef: > JPS(a)no.komm schreef: > > In message <1120808817.538631.149690(a)o13g2000cwo.googlegroups.com>, > > "mark" <mspan(a)mad.scientist.com> wrote: > > > > >I'm not going there (with you at least). > > > > Why not? > > > > >Bottom line is that when you > > >take raw information from the sensor, bayer sensors give you 1/3 color > > >info on each site, > > > > Actually. there is no loss of color coverage at all; the AA filter makes > > sure that a high percentage of light of all colors hitting anywhere on a > > particular point of the sensor plane reaches the three color sensors. > > So, in your opinion, on a bayer sensor the light that falls on a single > (say: green) input pixel also falls on its nabouring (red, blue) input > pixels? So in practice, a group of input pixels records the same 'ray > of light', allthough thay are in fact located on seperate positions on > the sensor. This is what the AA filter does? > > > It is just not possible to record color details above a certain > > frequency, which is a bit lower than the frequency of luminance detail > > recorded. We generally can't see color detail at those frequencies > > anyway, with normal viewing situations. > > And this luminance detail is recorded because the fraction of light > that does not belong to the 'high percentage of light of all colors' is > captured by the input pixels individually? > > > > > >foveon gives you all. Debayer software interpolates > > >the color info to create those humongous large files. > > > > Foveon does not give all. Foveon gives partial. Foveon has three color > > channels, but they are not particularly distinct, and the hue resolution > > can be pretty coarse, especially in the blue/green range. And the > > blue/green seems to often vary in large, obvious blotches, rather than > > in a manageable fine dither. > > > > Also, none of the current x3f-based cameras have proper aliasing, so > > they do not give you full spatial information. On the SD9, the capture > > has a very strong "snap-to-grid" effect, at the pixel level, and this > > makes the images look artificial to me. There are lots of artifacts > > that alighn themselves in horizontal and vertical lines, something that > > does not happen in real vision. > > > > As real vision is analogue? > > > If the Sigma cameras had proper AA filters, they would be better, IMO, > > but a lot of the people drawn to them might not like them, because they > > like the artificial sharpness of aliasing. > > > > You can reproduce the effect easily with any Bayer camera; upsample with > > bicubic to twice the dimensions of a Sigma image (4536*3024), and then > > downsize to 2268*1512 with Nearest Neighbor. Voila! Instant SD9 "pop" > > and staircase effects. > > tried that. no way. > > Honest questions! > > > -- > > > > <>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> > > John P Sheehy <JPS(a)no.komm> > > ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
From: mark on 26 Jul 2005 04:20 > > > <>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> > > > John P Sheehy <JPS(a)no.komm> > > > ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< Where are you?
From: mark on 12 Aug 2005 09:50 mark schreef: > > > > <>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> > > > > John P Sheehy <JPS(a)no.komm> > > > > ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< > > Where are you? John...
From: mark on 7 Sep 2005 10:16
mark schreef: > > > > <>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> > > > > John P Sheehy <J...(a)no.komm> > > > > ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< > Where are you? John... |