Why we have to use unsharp masking

[Motorsports Photo]

I thought someone beside myself might like a topic that didnt include "What lens should I use for_____"

Heres an interesting quote from "Evaluating Color Sampling Technologies" by Rudy Guttosch in the April 2005 issue of "Advanced Imaging"

Its a trade magazine that covers industrial and armed forces image technology.

"Most digital cameras utilize image sensors that use a mosaic of color filters to capture color. The filters are applied to each pixel through photolithography as a secondary process of sensor manufacturing. The Bayer mosaic pattern is the most popular design and consists of an alternating pattern of red, green and blue filters whereby 50% of the pixels have green filters, 25% have blue filters and 25% have red filters. Because each pixel captures only a single color, two out of three colors are not sampled at any one given point. Image sensors that use the Bayer mosaic pattern require the camera to perform color interpolation to estimate the missing colors from neighboring pixel data. For a red pixel location, for example, the green and blue data must be interpolated from green and blue neighbors because these values are not directly recorded.
Bayer pattern images are less sharp than they otherwise could be as a result of undersampling of color and post processing compensation that is applied to minimize visual artifacts. Color aliasing which causes white lines in the image, is a common side effect of Bayer pattern color sampling because of the lateral distribution of the color filters across the image sensor array. Optical or mathematical blurring is commonly used to reduce the presence of these artifacts, although by doing so, the overall sharpness of the image is significantly compromised."

I thought others might be interested in this technical, but understandable explanation.

-Pete

[steven]

It this problem that I keep hoping that the foveon technology catches on as they don't use a color mosaic filter and all three colors are take at every location. But I guess Canon (and all the camera makers) have too much money invested into their own indivudual tech to be interested in someone else's

[J Rabin]

That's an engineering discussion. It's nice information, like the Mars Rover beung a 1 Mp camera with big pixels. But few really care a hoot.
As long as resulting photos are attractive and meet people's quality expectations, and every camera generation improves. For practical purposes, current generation of dSLRs already exceeds the output quality of scanned 35mm film going to print. The engineers will continue to do their thing, thank gosh! J.

[ron chappel]

There seems to be some common missunderstandings about bayer sensors -even that article quoted hasn't got it exactly right (or more correctly doesn't tell the whole story)

If i recall rightly,foveon (and it's supporters) say that a foveon type chip has three times as much effective resolution as a comparable type bayer sensor.....but this is simply not the case.
And's it's very easily proven wrong by testing.

The reason a bayer sensor is not as blury as it's detractors say is that interpolation is applied to the *colour only* (not the actual pixels)
When one thinks about it,this is no big deal.Trying to guess what colour to apply to a given pixel is not too hard (how often do colours change throught a scene compared to the extremely dense pixel spacing?)

**Remember the relative lightness/darkness of each pixel is totally unaffected by the bayer matrix -and it's these bright /dark changes that show up as edges/sharpness**


Non Bayer sensors DO give more resolution per pixel but it's roughly in the order of 50% more,not the three times claimed by foveon
(that 50% was a guess,i havent studdied this subject for awhile and have forgotten most of the exact numbers-sorry)

[Jon]

Originally Posted by ron chappel

Non Bayer sensors DO give more resolution per pixel but it's roughly in the order of 50% more,not the three times claimed by foveon (that 50% was a guess,i havent studdied this subject for awhile and have forgotten most of the exact numbers-sorry)

Latest issue of Pop. Photo ran a sub-$2500 DSLR shootout that found the SD-10 had about 5 MP effective resolution, so about what you remember.

[adas]

What's even more interesting about it is that you can reach the full resolution of your camera by shooting BW subjects asnd ofcourse using an appropriate demosaicing algorithm. The question is where to find BW subjects elsewhere than shooting documents or the old grandfather's photographs?

[robertwgross]

Originally Posted by adas

The question is where to find BW subjects elsewhere than shooting documents or the old grandfather's photographs?

Chessboard with chess pieces.

---Bob Gross---

[Jesper]

I agree that the Foveon sensor idea sounds more attractive at first sight than it really is, if you think about it.

One other disadvantage of the Foveon sensor, which hasn't been mentioned yet here, is that it is more noisy than a Bayer sensor. Because the three pixels (R, G, B) are stacked upon each other, less light will reach the middle and bottom pixels, which means the signal has to be amplified more, resulting in more noise in the image.

[pierrot]

I do agree that Foveon technology looks more advanced at first glance.

The question is: what is the impact on real life photography? Do the images look sharper? Better defined? More contrasted? More true-to-life coloured? Is there a visible difference on enlargements?

Without examples, it's hard to make one's opinion. And besides of that, I wouldn't change my whole lenses collection just to be able to use a Sigma body, even if it is "better"...

[ssim]

What in god green earth did you jsut say. Yeah I got, didn't like and it hurt.

I just not that technically inclined and itf the test was next week I'm suire that I would flunk.