Go Back   Steve's Digicams Forums > Digital Cameras (Point and Shoot) > Panasonic / Leica

Reply
 
Thread Tools Search this Thread
Old Oct 28, 2005, 8:49 PM   #31
Senior Member
 
squirl033's Avatar
 
Join Date: Dec 2004
Posts: 6,720
Default

no, he said he could manipulate a 2MP image to make it look almost as good as a 4MP picture, and he wondered why. he didn'treduce 4MP image down to 2MP - he shot a photo at 2MP resolution, then upsampled and sharpened it. in any case, the result is that he discovered that the interpolated result was nearly as good as a shot taken at full 4MP resolution.

to show the difference between an untouched image shot at 2MP resolution and one taken at 5MP (i have an FZ20, not an FZ10/15), i submit the following:

here's a 5mp image, cropped to show an800x600 sample

http://img.photobucket.com/albums/v3.../barn5mpsm.jpg

and here's a 2MP original cropped to show the same image area, thenresized to the same 800x600"viewing size"

http://img.photobucket.com/albums/v3...rn2mpsmall.jpg

both images were cropped to show the same approximate area. the 5MP pic was just cropped, no resizing, while the 2MP image was cropped and then resized to the same 800x600 size, to simulate making the same size print from each, which is the only way to make a fair assessment of the relative image quality. clearly, the 2MP image is much worse - you can see pixels, and he clarity and detail simply aren't there.

most of us are totally unconcerned with Bayer filters (as opposed to Foveon), or color filter arrays or photodiode technology.we care about the image . if a 4MP imageis better (i.e. sharper and more detailed) than an 2MP image, that's what matters. if we later discover that both images only use half the pixels we think they do, that doesn't really matter, because it's proportional andthe images themselves don't change. the fact that my 5MP sensor may in reality only be using half the total number of pixels it says it is doesn't bother me, as long as i know it will make stellar 8x10 or 11x14 prints.


squirl033 is offline   Reply With Quote
Old Oct 28, 2005, 9:33 PM   #32
Member
 
Join Date: Oct 2005
Posts: 36
Default

Unless I misunderstood something, he shot one picture in 2 megapixel mode
and one in 4 megapixel mode, both with the same camera. As long as the
resampling algorithms are the same, resampling in the camera is no different
than resampling in Photoshop---unless JPEG artifacts are introduced into the
workflow. The fact is that a 4 megapixel image was resampled down to 2
megapixels. Just because 2 megapixel mode is selected doesn't mean the
CCD will sample an image AT 2 megapixels. It samples at the full resolution
to allow interpolation and edge copying to be done as accurately as possible.
Unless the selected output size is small enough, it is absolutely required that
all photodiodes be used. If the output size is small enough, not all photodiodes
need to be used. However, any high quality camera should use all of the photo-
diodes no matter what the final output size is in order to produce the most
accurate sample possible.
dimagez1 is offline   Reply With Quote
Old Oct 28, 2005, 9:41 PM   #33
Senior Member
 
squirl033's Avatar
 
Join Date: Dec 2004
Posts: 6,720
Default

dimagez1 wrote:
Quote:
Just because 2 megapixel mode is selected doesn't mean the
CCD will sample an image AT 2 megapixels. It samples at the full resolution
to allow interpolation and edge copying to be done as accurately as possible.
so in reality, he's comparing two 4MP images, and there's no point to this whole thread... :roll:

the images i posted in the preceding post are "real" 2MP and 5MP images, one taken with an FZ1v2 and the other with an FZ20. clearly, a "real" 2MP photo is vastly inferior to one taken using a 4-5MP sensor and playing games with internal sampling and resizing.




squirl033 is offline   Reply With Quote
Old Oct 28, 2005, 10:17 PM   #34
Member
 
Join Date: Oct 2005
Posts: 36
Default

Wow, you really misunderstood the point of this thread! Of course a 2 megapixel
camera will produce detail inferior to a 4 megapixel camera. The 2 megapixel camera
has the same Bayer pattern induced inaccuracies as the 4 megapixel camera. But,
because the 4 megapixel camera has more photodiodes to start with, it's obvious
it's resolution will be better. Neither camera will have the advertised resolution.
The whole point of this thread was to say that perhaps digital cameras don't really
have all of the resolution that they are advertised to have. That is unless you are
talking about a digital camera that uses the Foveon X3 sensor. Although it produces
a 3 megapixel image, it's detail resolving capability is equal to a 6 megapixel
Bayer pattern sensor. In some cases, it will actually outresolve the 6 megapixel
Bayer pattern sensor.
dimagez1 is offline   Reply With Quote
Old Oct 28, 2005, 11:10 PM   #35
Junior Member
 
rubinsky's Avatar
 
Join Date: Apr 2005
Posts: 14
Default

Hi Ted:-),

Wow Ted:-)! Way to go! Nice article. It think it puts a real damper ona misleadingword like "Megapixels". But as far as I know,and please correct me on this if i'm wrong, on myWindows machine (not including the output or process of the video card), each pixel is encodedto containa single varying color (24bits wide) in ordertoform a large multitude of 16,777,217million(derived from [2^n+ 1]=16,777,217 where n=24bits), which are stored in a 32 bit register. Needless to say the video card will convert this stuff to some form of RGBscheme for viewing purposes (unless someone out there has or will developed a single multi-coloredpixelformonitors and this is really besides the point).I mention this to denote that computers aren't storing Pixels in RGB schemes but as single units, which leads me to my next point.

My next point is that based on what you said about the "Bayer pattern CFA" (which I'm not against in any wayexcept for it's questionableability to renderpurity), and the fact that my camera'sresolution output being roughly half of the rated megapixels(~2million true pixels) thatit's rated at (4mp), thenthecamera's internal hardware is essentially interpolating, inflating and possibly sharpening roughly by2 million 24bit-widepixels into 4mp. It seems like such a large margin of error:shock:. So would I be correct to assume that all digital cameras that are based on the "Bayer pattern CFA" are being assigned inflated pixel output valuesto their specifications? I hope I'm not overreacting but does anyone else see this as cheating, or is it just me?:?:shock:

Nowbecause of the aforementioned "Bayer pattern CFA" by Ted, I question whether the so called "RAW Format" is trully raw at all because a set of algorithms would be needed to compose an image using the "Bayer pattern CFA". But after the algorithms are applied to the digital map of the CCD's contents, by definition, they are no longer trully "RAW". Ay-ya-yay!:?.

Now I need to collect my thoughts about my experiment. I alsoneed to take dimagez1's (Ted) new article and squirl033's advice into account before I can make a much needed revision to this my original post. I also want to create my own way of measuring true definition for any type of camera. I hope my peers are willing to test it with their camera becuase, well, I only have one camera(fz15).

By the way, thanks squirl033:-)



Hislipe:-),
What type interpolation method did you use to blow up the 2mp? I appears thatthe 2mp wasn't sharpenned properly. I took the 2mp and was able to sharpen just enough to resemble the 4mp, but because I don't have the originals i don't it will be a fair comparison.



Note to all: One of the points in this experiment is to see whether or not the camera is using some form of interpolation. It may be possible that the camera's hardware is doing some sort of interpolation and possibly sharpening (hardening) of its own. Soin the main part of the experiment I seek to know: Can a 2mp image be inflated (using bicubic sampling) and thensharpened at its best (using "Unharp Mask" filter in Photoshopto tweak the sharpness) tovery closely resemble the 4mp, look practically the same as the definition of a 4mp.But now this experiment will need to changeto someextentbased on Ted's new articleof "Bayer pattern CFA".


If the 2mpdoes look practically thesame asthe 4mp, it doesn't mean that Ican jump to conclusions of any kind. It just means to me that I willhave to developnew experiment tofind out why, with of couurse the much appreciated help and support of my peers if they are so willing,. As far as I'm concerned, theres no winning or losing in this post but only the search for some sort oftruth. Because, frankly I'm dead tired of being led around like a some stupid sheep by manufactures. Yes the fz15 is a pretty good camera, with minor problems and many limitations which are hustifiable at its base price(for the moment). But like any free person, I have the right towonder "why?".

I know that I cannot change the way things are, and in fact, I have no real interest in doing so. Everyday I see these "dead-end" specifications onmost products basicallythat lead to nowhere. So instead of making a futile attemptat changing the market, I aminstead, trying to put a point of real world referenceto these so called "Megapixels". Then, at least, I can weigh them withmy cognitive abilities and assign them real world attributes instead of pointless numbers and names.Most of you are already are aware of this stuff, so apologize in advange for spilling outmy common sense to you or for any blasphemous remarks I may make during the span of this post.

Mainly, I mentioned all of this because I get the felling that for practicing this experiment,I'm going to get burned at the stake, by the flames that may or may notsoon engulf this post. And I'm sure I will burn like any good witch, if it so is true. Not that I care for or give attentionflames, butitwould certainly be a crying shame to see this material go to waste. Especially now that Ted has been sogenerous asto take the time toinclude some of hisfactual knowlege.

-Ruben

rubinsky is offline   Reply With Quote
Old Oct 29, 2005, 12:20 AM   #36
Senior Member
 
squirl033's Avatar
 
Join Date: Dec 2004
Posts: 6,720
Default

Ruben,

Will your experiment be based on using a 4MP sensor to record a "2MP" image, or will you be attempting to make a true 2MP image look as good as a 4 by upsizing and resampling? from a purely academic standpoint, if your research shows that the FZ15 is really a 2MP sensor being internally upsampled to produce a 4MP file (which i very much doubt), i expect you'll throw a whole lot of people into a tizzy.

i have made "real" 2MP images, then upsized and sampled and processed them so that they look nearly as good as unretouched 4MP photos when printed at 8x10 size. i have also cropped 5MP files to leave only 3MP and then upsized and sharpened, and printed at 11x14, with amazing results. what does that prove? that pixel count often isn't as important as many people think, because modern resampling and processing software can do marvelous things.

in the end, the only real test that matters is what an image looks like. pixel count, when you get right down to it, is mostly a marketing tool, just like "watts per channel" in audio. it's what most consumers look for when they buy a camera, because they don't know enough to look for anything else. it's a quick and easy way for non-technical people to gauge the relative merits or potential image quality of one camera over another.

perhaps the real answer lies in making camera manufacturers redefine "pixel" as used in their marketing and technical literature, so that the pixel counts they claim match with Dimagez1's Bayer filter pixel definition.if all digicams using the Bayer filter (which, as i understand it, is almost all of them) turn out to produce only half the pixels they advertise, that may be a source of consternation to some, but as long as the companies all advertise them the same way,then pixel countstill just as useful for determining the relative potential capability of one camera against another. of course, all of those customers who thought they had 4MP cameras, and now realize they're reallyonly 2MP effectively, will be ticked off, but the smart ones will look at the pictures, and if what they thought was a 4MP image turns out to actuallybe only 2MP, but is still just as clear and sharp and detailed and as pleasing to the eye as it was before, will they care?frankly,i don't give a @#% if my camera has 1MP or 10; if it prints gorgeous images at the sizes i want, then that's all i ask of it. i'd much rather be making and enjoying photographs than worrying abouthow many pixels the sensor in my camerareally produces...

squirl033 is offline   Reply With Quote
Old Oct 29, 2005, 12:39 AM   #37
Member
 
Join Date: Oct 2005
Posts: 36
Default

Just remember that the advertised resolution IS necessary to maintain
a correct aspect ratio as well as every single bit of detail provided by
the sensor chip. It is all related to the physical positioning of the
photodiodes and their respective color. So, it isn't really cheating since
it is necessary.

If you were to take a true 3 megapixel image (such as one produced
by a Foveon chip) and resize it down to 1.5 megapixels, then you truly
would be losing detail that not even the most advanced interpolation
algorithms from 100 years into the future could "regenerate".

Remember this very important point: just because the typical digital
camera image really only has 1/2 the advertised resolution, that doesn't
mean you can reduce the size of the image to contain 1/2 as many pixels
without losing some amount of detail. Again, it is a technical matter
related to the physical placement of green photodiodes on the sensor
chip.
dimagez1 is offline   Reply With Quote
Old Oct 29, 2005, 1:15 AM   #38
Member
 
Join Date: Oct 2005
Posts: 36
Default

Squirl033,

This is meant to be a thread in which we can converse about the
technical aspect of resolution. Obviously, the actual content of the
picture is what is most important, but that's not what we are
discussing. Why do you seem to insist that resolution isn't a very
important aspect of photography? The more resolution there is,
the more lifelike a picture will look. This is usually one of the
goals in photography. You can't honestly tell me that resolution
isn't important to a lot of people out there. Some people may say
that it doesn't matter in the hopes that they will make themselves
feel better about their old camera. But sooner or later, they all
end up buying a higher resolution camera and then, at the very
least, admit that it is nice to have more detail. Resolution IS
important, that is why the market is so focused on increasing
the pixel count. That is also the reason behind HDTV. As long as
an image doesn't have as much detail as our eyes can see, then
there is room for improvement.

If more resolution isn't better, then why would a photographer
choose to buy the EOS 1Ds Mark II instead of a EOS 1D Mark II?
The 1Ds is much more expensive and actually performs worse
than the 1D Mark II. What's the 1Ds Mark II's biggest advantage?
It provides an image with twice as many pixels as the 1D Mark II.
dimagez1 is offline   Reply With Quote
Old Oct 29, 2005, 2:33 AM   #39
Senior Member
 
squirl033's Avatar
 
Join Date: Dec 2004
Posts: 6,720
Default

Dimage1z,

i don't know what gave you the impression that i said resolution isn't important. obviously it is. what do you think the 'actual content' of a digital photo IS? what i SAID was that resolution - megapixels - is primarily a marketing tool to sell cameras, and serves mainly to give non-technical buyers some basis for comparing the relative capabilities of their camera choices, much like EPAmileage estimates are used to guess the relative fuel economy of cars.we all know those estimates aren't what we'll really get, but weknowcars with higher numberswill generally get better mileage than cars with lower ones.

with digicams,most buyers equate more pixels with better images. that's true, but only up to a point, because once a certain threshold is passed, the human eye cannot detect any noticeable improvement in resolution.you alluded to this yourself, when you said that "As long as an image doesn't have as much detail as our eyes can see, then there is room for improvement."that's exactly my point. where that threshold is probably depends to a certain extent on the viewer's visual acuity, but i would guess thatat a standard print size like 8x10 (and all other things being equal, such as optics, noise, correct exposure, printer quality, etc.), most people couldn't tell the difference between a 5MP photo and one from a 7 or 8MP camera. that's because once pixel count passes about 175ppi, or perhaps 200 for those with exceptional vision, our eyescan'tdetectthe difference. in other words, at that print size, a 5 or 6MPimage DOES have as much detail as our eyes can see, so any improvement beyond that is not visible - or needed. obviously, at larger print sizes such as 16x20 or 20x30, more pixels are needed to provide the same level of detail,and a 10-12MP camera would be a definite plus if you plan to make prints that big.but atprint sizes smaller than 8x10 or 11x14, which is about as big as even most 'serious amateurs' make,the difference between a 5 or 6MP image and an 8MP or largerimage is, for all intents and purposes, negligible.it's physiology, not physics, and adding more pixels at that point yields no benefit.

an analogy: back in the 70's, manufacturers like Pioneer and Kenwood and Marantz all touted 'total harmonic distortion' and watts per channel as the differentiating features of their amplifiers. they bragged that they had 120 watts per channel, while the other guy only had 100 - as if a listener could tell the difference! and they boasted of THD figures of less than .01%, in spite of the fact that audio distortion has to be above .03% before it can be heard! once amplifiers passed that threshold, no one could hear any improvement, but people bought the better specs anyway, just because the salesman told them it was better. same thing with digital cameras. everybodywill getexcitedabout a 12MP camera, when in fact, in any normal print size, no one could tell whethera print was made from 12MP or 8. the eye simply cannot resolve details that fine, so to the viewer, all those extra megapixels are indistinguishable.

you asked why a photographer would buy a camera with more megapixels, even though it might not perform as well as one with fewer? you just proved my point... marketing! people think more is automatically better, just like the poor sod who spent $500 for that 120-watt Pioneer with .005% THD, and then couldn't understand why it didn't sound any better than his buddy's $200 amp with half the power and twice the distortion.all this means is, as a manufacturer,you can claim any darn thing you want, as long as whatever youclaim is beyond the ability ofyour customer to see or hear, and no one without an engineering lab will be able to prove you wrong. your customer will surely never know the difference.

i'll bow out of this thread now, and leave you and Ruben to your esoterictechnical discussion of the physics of digicamsensors and the merits - or lack thereof -of Bayer versus Foveon filters...


squirl033 is offline   Reply With Quote
Old Oct 29, 2005, 5:34 AM   #40
Junior Member
 
rubinsky's Avatar
 
Join Date: Apr 2005
Posts: 14
Default

Dam'n you know, I never thought I'd miss a squirrel, but after you said you were not coming back, Irealized that the only imageI have of you is a squirrel. So now i miss a squirrel

Esoteric? I know some of this postcan get a little strange at times but I don't how it is mystical?:?

Here is some strange for you:The only way to trully improve sound quality is to increase the amplifier's output damping factor to infinity(impossible now) by producingbipolar mosfets with superconductor like characteristics (impossible now), and eliminating input impedence thereby cutting noise.And even if someone doesfigure out how to do this, they will also have to produce speakers with superconductor coils. Although amplifiers have great performancesince the olddays, they all lack the most critical ingredient to keep speakers from going astray and cycling incorrectly and out of sync. Solution: a perfect, 0-ohm active damping factor with a zero seconddelay "feedback-error correctingnetwork". Sound impossible? At the moment, yes. How do i know this? I designed amplifiers for about2 years and I studied electronics for 4.

But you are right squirrel, ampshave'ntbeen ableto improve sound quality during tha span of that time (at least not by much). At this time,the issuerest solely on the speakers because of the current technology doesnt allow power transistors to be more efficient.

Moral of this story: I feel kinda sorry for all the poorsouls that wasted their hard earned money on misleading specifications. And while I can't save them, maybe we can help rescue a few here by, at least,partially removingthose dreadedad-catching marketing blindfolds.



Sorry to hear you're notcoming back squirl033. Hope you're not upset...There goes my moral

I'll will try to work on the resolution experiment tomorrow. Sorry for the delay.



nice foveon pic:

http://www.foveon.com/files/H.Germ_W...rrot/large.jpg
rubinsky is offline   Reply With Quote
 
Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



All times are GMT -5. The time now is 9:57 AM.