Why only 4.6MP?
in [Digital Cameras]
|
Prev: Processing Program
Next: curanzia versicherung, berufsunf�higkeits versicherung, berufsunf�higkeitsversicherung f�r selbst�ndige, berufsunf�higkeitsversicherung preis, versicherung vergleich,
From: nospam on 1 Mar 2010 13:49 In article <XMTin.38825$K81.16877(a)newsfe18.iad>, Martin Brown <|||newspam|||@nezumi.demon.co.uk> wrote: > A quick glance but I could not see any suitable image that combined > being tack sharp with having saturated red with black lines on it > thinning to single pixel. Situations where it really matters are rare. not even the resolution charts? <http://www.dpreview.com/reviews/sigmadp1/page19.asp> <http://www.dpreview.com/reviews/sigmadp1/page20.asp> > Chroma subsampling exploits a feature of our eyes limited colour > resolution as compared to brightness. which is why the entire foveon concept is a solution in search of a problem.
From: Martin Brown on 1 Mar 2010 13:22 Pete wrote: > Martin Brown wrote: >> You have to be close enough to see individual pixels (or at least be able >> to tell the difference between one pixel black and one saturated red vs >> two pixels half brightness red. Fine detail like black veins on a red >> flower like a poppy will show significant artefacting. But most of it is >> down to the JPEG decoder and 2x1 chroma subsampling no the sensor. >> >> The win lose for Bayer vs Foveon is extremely sensitive to the material >> being photographed. The testing case is fully reolved fine black liens on >> a saturated red background. The only time I have had real problems with >> Bayer sensors was photographing in monochromatic red light. >> >> Obviously operating with half resolution in both dimensions. The effects >> and limitations of the demosaic were obvious - although even then *most* >> of the defects in the final image came from systematic faults in the >> publicly available JPEG decoders as currently implemented. >> >> If someone wants to photograph a subject that is challenging for Bayer >> sensors with both a good 8Mpixel Bayer camera and a Foveon I will decode >> the Bayer image to demonstrate that most of the benefit Foveon devotees >> are claiming is actually down to 1:1:1 chroma subsampling and a fault in >> the JPEG decoder spec for chroma subsampling as presently implemented. > > Martin, my question is not related to the subject of this thread, but what > you've written may explain an ongoing problem I've had with some software. It might be the software. The only package I know of where the default settings will rapidly degrade an image saved as JPEG at the same nominal quality settings is PSPro v8. Their codecs chroma subsampling is broken. It is more or less ok at 4:4:4 no subsampling. > > I'd been assuming that the rapid degradation with JPEG load-then-save > cycles, at highest quality setting, was caused by a level mapping error when > this software converts JPEG to linear and vice versa. My simple tests use a > lightly saturated green/magenta diagonal gradient and combinations of JPEG > and 16-bit TIFF input and output: it appears that the software has a small > error on linear to JPEG conversion, but most of the error is when reading a > JPEG and converting to linear. I set the colour space to sRGB for image, > working space, and monitor, to remove as many variables as possible. Strange - the image should drift slightly on the first couple of saves and then converge to a local attractor. I'd be interested looking at an image where degradation is cumulative for repeated load save at constant highest quality settings. That is unusual not to say abnormal behaviour. My strange looking email address is valid without modification. > My extensive Web searches have not revealed a tangible explanation, probably > because my keywords are limited by what I know. I'm attempting to learn > something new therefore I don't know suitable keywords. Any help would be > greatly appreciated. It should not be doing what you are observing. If you change contrast, black level or gamma correct then all bets are off, but a basic load save should converge relatively quickly on a self consistent solution. I did some work on this a while back. See http://www.nezumi.demon.co.uk/photo/jpeg/jpeg.htm And in particular the link on generational losses. Regards, Martin Brown
From: Martin Brown on 1 Mar 2010 13:34 nospam wrote: > In article <OqLin.882$cp7.294(a)newsfe23.iad>, Martin Brown > <|||newspam|||@nezumi.demon.co.uk> wrote: > >> If someone wants to photograph a subject that is challenging for Bayer >> sensors with both a good 8Mpixel Bayer camera and a Foveon I will decode >> the Bayer image to demonstrate that most of the benefit Foveon devotees >> are claiming is actually down to 1:1:1 chroma subsampling and a fault in >> the JPEG decoder spec for chroma subsampling as presently implemented. > > can you use the samples from dpreview or another review? A quick glance but I could not see any suitable image that combined being tack sharp with having saturated red with black lines on it thinning to single pixel. Situations where it really matters are rare. Closest was their still life the C M Y colour spots at the top show some visible artefacting as do the black lines on the small blue globe. They are not as black as they should be because chroma bleed is affecting them. If someone wants to do it the radial resolution charts one white, black and the others blue, black and red black will show what I mean. I would actually be interested to see if the Foveon can beat the 8Mpixel clasic Bayer CCD. My instinct is that it cannot. Real images seldom have the right sort of detail to make the Bayer mask important even pixel peeping. Most of the loss is due to decoder faults not demosaicing. Chroma subsampling exploits a feature of our eyes limited colour resolution as compared to brightness. Regards, Martin Brown
From: Alfred Molon on 1 Mar 2010 14:24 In article <010320100310505117%nospam(a)nospam.invalid>, nospam(a)nospam.invalid says... > yes, but the foveon has an *additional* need to keep the three layers > consistent. You are making some assumptions about things you do not know. There could very well be no additional step. -- Alfred Molon ------------------------------ Olympus E-series DSLRs and micro 4/3 forum at http://tech.groups.yahoo.com/group/MyOlympus/ http://myolympus.org/ photo sharing site
From: Alfred Molon on 1 Mar 2010 14:26
In article <hmfrir$o9c$1(a)news.eternal-september.org>, david- taylor(a)blueyonder.co.uk.invalid says... > >> As the pixel size in a 10MP Foveon camera would be smaller than the 4 > >> Bayer pixels in a 12-15MP DSLR, not larger, it could place higher > >> demands > >> on the lenses, not less, and diffraction could be more of a problem. > > > > Same photosite size, but higher effective resolution. The pixels in a > > bayer camera would have to be roughly sqrt(1.5) times smaller to equal > > the resolution of a full colour sensor. > > > > Example (pixel size): > > Full-colour: 8 micrometers > > Bayer with equivalent resolution: 8/1.224= 6.5 micrometer > > > > The Bayer pixels are smaller because you need more of them. > > > > That means, with a full colour sensor you get an effective resolution > > increase without making the pixels smaller. Fewer demands on the lens, > > fewer diffraction problems. > > False, if the low-pass filter does the correct job for Bayer pixels. Did you even read what I wrote? Your reply has no connection at all with my post. -- Alfred Molon ------------------------------ Olympus E-series DSLRs and micro 4/3 forum at http://tech.groups.yahoo.com/group/MyOlympus/ http://myolympus.org/ photo sharing site |