Am I oversharpening
in [Digital Cameras]
|
From: JPS on 1 Jun 2005 21:29 In message <1117594696.335653.87000(a)z14g2000cwz.googlegroups.com>, george_preddy(a)yahoo.com wrote: >When a Bayer camera says "8MP" that means "8MP MONOCHROME", not 8M full >color pixels. It means neither. It means that 8M samples are taken, in three different color bands, at a total of 8M 2D locations. This is not as accurate as taking 24M samples in 8M 2D locations, but is much more detailed than taking 8M samples in 2.66M 2D locations. There is absolutely no reason why the three color samples have to be taken at the same 2D location. -- <>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS(a)no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>><
From: Ken Tough on 2 Jun 2005 04:16 JPS(a)no.komm wrote: >george_preddy(a)yahoo.com wrote: >>When a Bayer camera says "8MP" that means "8MP MONOCHROME", not 8M full >>color pixels. > >It means neither. It means that 8M samples are taken, in three >different color bands, at a total of 8M 2D locations. This is not as >accurate as taking 24M samples in 8M 2D locations, but is much more >detailed than taking 8M samples in 2.66M 2D locations. > >There is absolutely no reason why the three color samples have to be >taken at the same 2D location. In fact, one should consider the characteristics of typical images to see which technique is best. In standard images, colours changes gradually or at sharp edges, not in random 1 pixel patterns. That means the interpolation provided by the Bayer filter will give better resolution and more information where it's needed. It's similar to things like image compression applying to typical images but being useless on completely random noise patterns. This "George Preddy" seems utterly desperate to rationalise his purchase of a foveon sensor camera. He'd probably expend less energy pulling his hair out if he just dumps it now and gets a standard one before it's too late. But then again, he might be the type who still drives a wankel engine car. -- Ken Tough
From: Black Nikon on 4 Jun 2005 20:29 With the first photo's your are indeed oversharping, but the rest is okay. Eric "Celtic Boar" <extraneous(a)charter.net> schreef in bericht news:PUsme.17445$cP2.6248(a)fe06.lga... > Please take a look at the attached link. I am still trying to get the hang > of this Unsharp Mask Thing. Are these oversharpened. > > Canon 20D - Raw - 75-300 IS Zoom > > Thanks. > > http://spaces.msn.com/members/fleetingglimpse/PersonalSpace.aspx?_c11_PhotoAlbum_spaHandler=TWljcm9zb2Z0LlNwYWNlcy5XZWIuUGFydHMuUGhvdG9BbGJ1bS5GdWxsTW9kZUNvbnRyb2xsZXI%24&_c11_PhotoAlbum_spaFolderID=cns!1pioagCJB9TmXNBFRp_AlULA!741&_c=PhotoAlbum > > ~Rikk > > website: www.fleetingglimpse.com > blog: http://spaces.msn.com/members/fleetingglmpse/ >
From: george_preddy on 4 Jun 2005 21:23 Ken Tough wrote: > JPS(a)no.komm wrote: > > >george_preddy(a)yahoo.com wrote: > >>When a Bayer camera says "8MP" that means "8MP MONOCHROME", not 8M full > >>color pixels. > > > >It means neither. It means that 8M samples are taken, in three > >different color bands, at a total of 8M 2D locations. This is not as > >accurate as taking 24M samples in 8M 2D locations, but is much more > >detailed than taking 8M samples in 2.66M 2D locations. > > > >There is absolutely no reason why the three color samples have to be > >taken at the same 2D location. The point is they have to be taken. Bayer hardly takes any color samples given their phony monochrome MP ratings. Listing monochrome MPs is a scam. There are NO WHERE NEAR enough color samples taken with a Bayer sensor to support the advertised color MPs. > That means the interpolation provided by the Bayer filter will give > better resolution and more information where it's needed. Interpolation never adds optical resolution. Try it for yourself, download a satellite picture of the world and interpolatively upscale it using Photshop. Make it a bigger and bigger image until you can actually see yourself smiling in the picture. Obviously, inserting interpolated digital placeholders into an image (upscaling) never improves optical resolution. Not an iota, ever. Bayer DSLRs are characteristically blurry and soft because the true optical resolution is the resolution of each, individual, tiny, subdivided, R, G, or B exposure, not the 400% interpolatively upscaled version, which is required to topologically align and overlay the three, due to their imbalanced size.
From: JPS on 5 Jun 2005 06:45
In message <1117934606.605577.297030(a)g47g2000cwa.googlegroups.com>, george_preddy(a)yahoo.com wrote: >The point is they have to be taken. Bayer hardly takes any color >samples given their phony monochrome MP ratings. Listing monochrome >MPs is a scam. There are NO WHERE NEAR enough color samples taken with >a Bayer sensor to support the advertised color MPs. They're not advertised as recording "color megapixels". >Interpolation never adds optical resolution. > >Try it for yourself, download a satellite picture of the world and >interpolatively upscale it using Photshop. Make it a bigger and bigger >image until you can actually see yourself smiling in the picture. Don't be ridiculous. Everyone arguing with you knows that you can't get more detail by upscaling a bitmap. De-mosaicing Bayer data is nothing like upscaling a bitmap, because some aspect of every pixel output is real and measured. If you quadrupled *that* (like 25.2MP from a 10D), then *that* would add no detail. Rendering it at 6.3MP does not "upscale" the image. Making a 14MP image in SPP, of course, *IS* the same thing as your satellite image example. >Obviously, inserting interpolated digital placeholders into an image >(upscaling) never improves optical resolution. Not an iota, ever. That is not what is happening. Why can't you get anything right? >Bayer DSLRs are characteristically blurry and soft because the true >optical resolution is the resolution of each, individual, tiny, >subdivided, R, G, or B exposure, not the 400% interpolatively upscaled >version, which is required to topologically align and overlay the >three, due to their imbalanced size. The luminance, the most significant part of the capture, is not upscaled at all. Red and blue resolution are upscaled to 200%, and green resolution is upscaled 70.7%. -- <>>< ><<> ><<> <>>< ><<> <>>< <>>< ><<> John P Sheehy <JPS(a)no.komm> ><<> <>>< <>>< ><<> <>>< ><<> ><<> <>>< |