Have we all been duped?

[mjferron]

<p>Roger your comment reminded me to finish it up so I did. My first impression was how scratched the color negatives were. Many spots and hair-like flaws as well. (How do they do that and why don't they show up on the photo CD? My B&W negs I process at home are very clean.) Anyway I scanned them on my Scan Dual 4 at the max 3200 resolution and down sampled to 5x7 at 300 dpi. Cleaned up the flaws, got rid of a colorcast in the whites and sharpened to taste. Verdict, due to the colorcast correction I liked my print from my Epson 2400 better than the Kodak print I recieived. Also the Moab Lasel Luster paper I used had a nicer looking texture and finish in my humble opinion. Verdict #2 In volume print film is only worth it if someone else is doing the work for you. Exceptions would be a photo that's really worth the time and bother of scanning. I find Reala's colors very pleasing.</p>

[danielleetaylor]

<p><em>The current practice among camera makers in listing megapixels is somewhat deceptive with respect to describing the sensors. When they say 10 megapixels they do not mean that there are 10 million green sensors, 10 millions red sensors, and 10 million blue sensors.</em></p>

<p>I know this, which should be apparent from my technical descriptions above.</p>

<p><em>They mean that (if it is a Bayers sensor) there are 5 million green sensors, 2.5 million red sensors, and 2.5 million blue sensors. The three types of sensor are interleaved on the chip. They produce three images, a green image sampled at 5 megapixel, a red image sampled at 2.5 megapixel, and a blue image sampled at 2.5 megapixel.</em></p>

<p>No, the sensor produces a single image. Luminance is sampled at every point. Chrominance is biased at each point in a known fashion (R, G, or B). By looking at the output from neighboring pixels and knowing the bias of each point, a computer can reconstruct the correct chroma value for each pixel.</p>

<p>One key thing you have to realize is that we don't generally encounter truly pure red, green, or blue in the real world. If you photographed a specially designed chart with patterns in pure colors that were completely filtered by the Bayer mask then you might be able to show a significant resolution loss and/or significant color misinterpretation. I'm not sure how you would produce such a chart. For that matter I don't know if it's a valid assumption that Bayer masks act as hard and fast chroma filters. Does a green cell completely block wavelengths outside of its range, or only attenuate them? Even very expensive filters do not cut like a knife, so what's the wavelength overlap?</p>

<p>At any rate, I know I've seen photographs of line charts printed in primary colors and the Bayer sensor DSLR still managed to resolve close to its predicted Nyquist limit. (I'm trying to find the reference.) This flies in the face of the idea that a Bayer sensor is only "half resolution."</p>

<p><em>The Nyquist sampling theorem prevents the green image from having a resulotion any higher than that of a pure 5 megapixel green sensor.</em></p>

<p>Again, Nyquist has nothing to do with Bayer or with color. Nyquist describes how to convert between digital sampling rates (in this case pixel density) and analog signal frequency (in this case lpmm). You didn't even use Nyquist theory correctly in the sentence above.</p>

<p>Bayer does not limit a 10 MP camera to a "green resolution" of 5 MP. For one, detail perception is determined primarily by luminance, not chrominance, so it's not meaningful to discuss something like "green resolution." This has been known and put to good use in color video since the first color TVs to reduce the bandwidth necessary to transmit a full color image. For two, this implies that the camera cannot determine green chroma values for half the pixels. If that were true then a photograph of a leaf would be horribly noisy with red and blue pixels forming half of the leaf. That's not the case. Bayer looks at the luminance data for all the surrounding pixels and, knowing their chroma biases, can very effectively determine the correct chroma value for a given pixel. But every photo site samples luminance and contributes to resolution. Bayer only fails to resolve a detail or interpret a color under very narrow and specific circumstances. It simply does not "lose" half the resolution of the sensor array, and this can be easily demonstrated with any DSLR and a line chart.</p>

<p><em>Now another word about the interpolation scheme: It may be that some schemes try to fill in the missing information in the green array by using information from the red and and blue arrays. The problem is that the red and blue arrays do not contain green information.</em></p>

<p>This is only a problem in a theoretical world where only pure red, green, and blue are encountered and completely filtered by color masks. Real photographs do not meet this criterea.</p>

[mjferron]

<p>PS I took the SD card from my Canon G9 into the local CVS and had a few prints made just to see what they would look like. The machine only offered 4x6's without waiting an hour so I used that option. I was quite impressed with those as well. Much faster then editing and printing at home. Not for the fine art photographer but perfect for the family album or prints to pass out to friends.</p>

[don_e]

<p>Film doesn't have rounding errors.</p>

 

[alan_rockwood]

<p>Dear Daniel,</p>

<p>I probably know as much about the Nyquist sampling theorem as almost anyone here. (I am sure there are a few exceptions, but likely not many.) I have published several peer reviewed papers in the scientific literature that used the Fourier transform, and in some of them I had to deal with the Nyquist limit, and also the somewhat related issue of the convolution theorem, so I am pretty familiar with the general concepts involved.</p>

<p>On the issue of luminance, I will not dispute the notion that luminance resolution could, in some cases, be unexpectedly high. (By "unexpectedly" I mean taken in the context of my discussions above.) For example, in a perfectly black and white image there may be little or no luminance distinction between the red, green, and blue sensors, and the luminance resolution of a so-called 10 megapixel Bayer sensor might approach that of an ideal 10 megapixel sensor. By "10 megapixel ideal sensor" I mean one which detects red and green and blue at each of 10 megapixels. However, there ain't no free lunch. For example, I have not worked out the details, but I suspect that you may end up with lower choma resolution in those cases.</p>

<p>It can easily be shown that a 10 megapixel Bayer sensor is not equivalent to a 10 megapixel sensor that samples r and g and b at each pixel by considering a special case. Consider an object composed of cyan, magenta, and yellow dots. The construct an object such that the dot-array of the object is in the same pattern as a Bayer sensor, according to the following specification. There are 5 megapixels of magenta dots. There are 2.5 megapixels of cyan dots. There are 2.5 megapixels of yellow dots. They are geometrically arranged in the same pattern as an RGB Bayer sensor, but with Magenta replacing the green position, cyan replacing the red position, and yellow replacing the blue position.</p>

<p>Now, select an magnification and orientation of the camera such that the spacing of the dots of the object are imaged onto the sensor in a 1:1 relationship defined as follows: The each magenta dot of the object is imaged onto a green sensor of the Bayer array. Each cyan dot of the object is imaged onto a red sensor. Each Yellow dot is imaged onto a blue sensor. The bayer sensor in this case will produce a perfectly black image at every point. On the other hand, a sensor that separately detects r g and blue at every point on the array will image this as an interleaved set of magenta, cyan, and yellow dots. Clearly, the two types of sensor arrays are not equivalent.</p>

<p>The example I give above is somewhat contrived because it the object does not satisfy the Nyquist limit of the sensor, but it is sufficient to prove that the two types of sensor are not equivalent. However, one can easily construct an object that does satisfy the Nyquist limit that will give a perfectly black image on a Bayer sensor but a series of resolved lines on a true rgb sensor. By "true rgb sensor" I mean one that simultaneously and separately detects r and g and b at each sensor point. I will point the way by discussing an object in which the magenta dots are modulated in intensity according to the following relative values: 1.0, 0.5, 0, 0.5, and 1.0 as you step along the array in the x-direction. This image has an overall value of 0.5 with an approximately sinusoidal pattern superimposed on it with a spatial period of 4 points, a spatial frequency that is 2X below the nyquist limit. A true RGB sensor will record the Magenta image correctly (a sinusoidal magenta pattern). However, a Bayer sensor of the same point spacing may record the Magenta image as perfectly black. One can modulate the Cyan and Yellow parts of the object by an analogous pattern. They may also be imaged as perfect black on a Bayer sensor, but a sinusoidal pattern of the appropriate colors on a true rgb sensor. There is no amount of image processing that you could apply to the (perfectly black) Bayer image in this case which would restore the true image, but the true rgb sensor gives a true image. You may change the phasing of the system (shift the image a bit). The true rgb sensor will continue to generate a true image, regardless of the phasing, but the Bayer sensor will show various distortions of the image, depending on how the object is phased with respect to the image space.</p>

<p>As I indicated in an earlier post, interpolation may give a good representation of the true image in some cases, but as shown in the examples discussed above it cannot be counted on to be correct. Your last post has already come close to acknowledging this by waffling on the topic when you say "Bayer only fails to resolve a detail or interpret a color under very narrow and specific circumstances." By contrast, the true rgb sensor gives a true representation of the image in all cases that satisfy the Nyquist sampling theorem.</p>

<p>Alan</p>

[alan_rockwood]

<p>One more thing. There is a subtle flaw in my description of the arrays and whether they satisfy the Nyquist sampling theorem. See if you can spot it.</p>

<p>However, the discussion is still sufficient to show that a Bayer array is not equivalent to a true rgb array. Also, it is possible to construct other examples showing that the two arrays are not equivalent which are not subject to the subtle flaw referred to above.</p>

[alan_rockwood]

<p>Here is an interesting Wikipedia link that discusses the demosaicing process, along with an example.</p>

[mauro_franic]

<p>To add to Alan's comments, no digital camera I ever tested has reached even 2/3rds of the resolution rated in its megapixel count. This is, even the demosaicing logic oriented to boost detail-luminance resolution at the expense of chroma resolution, can't pass this mark. Check any resolution test from any camera at DPreview to validate.</p>

<p>Yes, a 10 megapixel count has effectively 10 megapixels. People are duped (in many cases they dup themselves even if the manufacturer is honest) into reading it as 10 megapixels OF RESOLUTION - which is not.</p>

<p>Also, comments from people arguing -just to feel better with his dismissal of film- that a 4000dpi Nikon Coolscan captures all the information available in Velvia for normal contrast scenes found in the real world: Have no experience, no velvia shots in front of them, microscope, scanners or light tables, etc, </p>

<p>The results I posted several weeks ago of cans of tomato, spices, flowers and resolution charts using 35mm Velvia, TMX and Ektar and a Canon 40D shown massive difference in between just the scans of the films and the digital picture. When looked on the light table, this difference was like comparing my camera cell phone with the 40D.</p>

[danielleetaylor]

<p><em>I probably know as much about the Nyquist sampling theorem as almost anyone here.</em><br>

<br /> Alan, with due respect, you have misused and misapplied the theorem in this thread. Or at least my interpretation of your posts leads me to believe you have.<br>

<br /> <em>It can easily be shown that a 10 megapixel Bayer sensor is not equivalent to a 10 megapixel sensor that samples r and g and b at each pixel by considering a special case.</em><br>

<br /> I never said it was. But you claimed a 10 MP Bayer sensor is really a 5 MP sensor. This is false. A 10 MP sensor is a 10 MP sensor. And the Bayer filter simply does not cause a loss of 50% of the resolving power of a 10 MP array.<br>

<br /> <em>Consider an object composed of cyan, magenta, and yellow dots.</em><br>

<br /> No thanks. I'm not into complex theoreticals with no observational data to back them up. (I've had my fill of mental games from another unrelated thread.) I can imagine a specially designed test that could exploit the Bayer pattern to cause a large loss of resolving power. But the world we photograph is not composed of such theoretical (and extremely difficult to build/align/use) test charts, so it's pointless to appeal to such a mental construct and argue that 10 MP sensors are really 5 MP sensors.</p>

[danielleetaylor]

<p><em>To add to Alan's comments, no digital camera I ever tested has reached even 2/3rds of the resolution rated in its megapixel count...Check any resolution test from any camera at DPreview to validate.</em></p>

<p>Do you understand how DPreview's chart relates to pixel density and Nyquist? Because I just checked both the Canon 50D and 40D and they resolve well over 90% of the expected Nyquist limit. This is in line with Norman Koren's tests.</p>

<p><em>Yes, a 10 megapixel count has effectively 10 megapixels. People are duped (in many cases they dup themselves even if the manufacturer is honest) into reading it as 10 megapixels OF RESOLUTION - which is not.</em></p>

<p>You're mixing terms and definitions just like Alan. MP is a measure of resolution and 10 MP = 10 MP. There are other measures of resolution which are tied to subject matter versus machine design. One 10 MP sensor design might resolve more lpmm than another 10 MP sensor design, but this does not mean the first is not really 10 MP.</p>

<p>It's like racing two 300 hp cars and saying the slower one is not "really 300 hp." Yes it is. But how hp translates into maxium speed involves other variables.</p>

<p><em>Also, comments from people arguing -just to feel better with his dismissal of film- that a 4000dpi Nikon Coolscan captures all the information available in Velvia for normal contrast scenes found in the real world: Have no experience, no velvia shots in front of them, microscope, scanners or light tables, etc,</em></p>

<p>I've shot plenty of Velvia, scanned it and had it scanned, and viewed it under a microscope. And you are way out of line with your last sentence. If you want to have a friendly conversation, have one. If not, leave.</p>

<p><em>The results I posted several weeks ago of cans of tomato, spices, flowers and resolution charts using 35mm Velvia, TMX and Ektar and a Canon 40D shown massive difference in between just the scans of the films and the digital picture. When looked on the light table, this difference was like comparing my camera cell phone with the 40D.</em></p>

<p>I never participated in those threads, but at first glance something is fishy with your tests. I've never seen a DSLR image so pixelated. While 35mm Ektar and TMX should out resolve an APS sensor 40D, some aspect of your test is really skewing the results and increasing the gap. I don't have time to review old threads and try and figure out what that might be.</p>

[richard_p]

<p>If I remember correctly, you did participate in some discussion. Mauro kindly ask you to present your experience related to digital SLR, scanner, and microscope. Just want to set the record straight.</p>

<p>Richard </p>

[mauro_franic]

<p>Daniel, I did ask you to present your proof and you never did.<br>

<br /> In my case though, I shared the raw files and the actual film for people to verify.</p>

<p>"<em>Also, comments from people arguing -just to feel better with his dismissal of film- that a 4000dpi Nikon Coolscan captures all the information available in Velvia for normal contrast scenes found in the real world: Have no experience, no velvia shots in front of them, microscope, scanners or light tables, etc,</em> <br /> <em></em> I've shot plenty of Velvia, scanned it and had it scanned, and viewed it under a microscope. And you are way out of line with your last sentence. If you want to have a friendly conversation, have one. If not, leave."</p>

<p>Daniel, are you saying that you don't see more information under the microscope than with the Coolscan? or are you agreeing that Velvia has more information than the Coolscan captures? Would you mind posting your results either way?</p>

[Dave Luttmann]

<p>Odd. I have no problem seeing more detail under a magnifier or microscope than in a 4000ppi scan. Of course, if the cameras optics are a weak link, then one won't. With good optics, I can easily see more detail in an Imacon scan at 6300ppi than I can in the 4000ppi scan. The fact that many of us can means that those who can't must have weak optics....<br>

As well, if a 10mp Bayer sensor IS INDEED 10mp in effective resolution, then I'd like to hear why monochrome sensors outresolve Bayers sensors by a good margin. There was a good reason the old Kodak DCS 760M outresolved the 6mp Bayer sensor.....some of us know why.</p>

[VKinCA]

<p>Digital cameras are helping to save the computer industry. And since most photogs are such gadgeteers and always looking for a more magical magic bullet, the money will keep rolling in for companies offering digi-goods. I'm happy with my 2004 Fuji 8M FinePix rangefinder for casual and family color shots, but for serious B&W landscape work, I'll bring the Linhof Tech IV or Mamiya RB67-S and process / print the negs myself to Silver Gelatin Fiber Paper.</p>

[alan_rockwood]

<p>Daniel,</p>

<p>You said "MP is a measure of resolution and 10 MP = 10 MP. "</p>

<p>From your comment I am guessing that you also believe in upsampling as a way to increase resolution. It does not. It only increases the number of points in the image.</p>

<p>The number of megapixels in a final image sets an upper limit on the resolution an image can have, but other than setting an upper limit it has virtually no relationship to resolution. Furthermore, if one digital image is derived from another (such as by demosaicing of a Bayer image, essentially a special form of upsampling) the upper limit of resolution in the final image is determined by the weakest link in the chain, i.e. the image that had the fewest sampling points.</p>

[jeff_z.]

<p>Michael,</p>

<p>These comments resonate with me: "..My thoughts behind the post were about how easy it is to drop off a roll of print film and get a nice set of prints and a photo CD handed back. All done, very little fuss.." And, "..I just think print film is becoming a forgotten tool in the bag.."</p>

<p>I've never strayed from film, but got very caught up in scanning and printing on an Epson 2200. Then about two years ago, I shot a few rolls of print film of a kids' athletic field day. Had them developed by a good local lab, 4x6 inch prints.. It was an epithany, as I was reminded of what excellent results 35mm film developed and printed competently, can produce! And of course, how easy and even economical this is when you count your time and materials.</p>

<p>When we see how many seemingly endless hoops people will jump through in trying to make the argument for digital capture, and the kind of easily obtainable results from what you originally mention, it's no contest, at least for many of my objectives. Sure, if I want to make an occasional enlargement of a special print for framing, I'll scan and print..</p>

<p>Also, after years of shooting E-6 and color print, I just tried my first couple of rolls of B&W with a rangefinder. Kodak T-Max 400. Incredible stuff! I've never had so much pure fun with photography as with this combination.</p>

<br />

<p> </p>

[Dave Luttmann]

<p>Careful Jeff, once you start B&W in a rangefinder....there's no turning back. My DSLRs come out a lot less now that I work with my rangefinder and 4x5.</p>

<p>Film.....is real!</p>

[mjferron]

<p>Jeff my first love is B&W film developed at home. Mostly scanned and every once in a while I break out the enlarger. Usually on an old calssic camera as well. Right now I have a Rolleicord VA and a Rollei 35s. As Dave says above. It's real.</p>

[jeff_z.]

<p>Dave, Interesting! Yes, this seems absolutely addicting!</p>

<p>Michael, I may well attempt to learn to do this at home if my quantity increases a lot. I've gotten outstanding results from Philadelphia Photographic (mail order), though, and their prices are pretty decent.</p>

<p>There is definitely something special about this photo genre:)! Do you ever scan and print BO? I have the Epson 2200 and a good dedicated film scanner.</p>

[roman_p]

<p >Michael,</p>

<p > </p>

<p >This pic is on Fuji Super HQ (4 bucks per pack of 4) with Nikon F601M ($30 on e-Bay) and processed in Wal-Mart. I adjusted levels and applied Unsharp mask using my 6yr old laptop and PhotoShop V.6 which I got for free. I spent 5 min total time for editing. A friend of my lately spend over $700 to accrue his D40 kit. His is a low level armature and when he came in local Ritz Camera store, the sale person told him that the digital is much cheaper, convenient, reliable (!), revolutionary innovated and it’s only way to go in our bright human future (well, there’s no film cameras to compare anyhow!). After knowing how much money and time I spend to create those pics, my friend thinks he was duped. I don’t blame the sale person much. If he was trying to say something appraising film photography he would get fired immediately. And not only this salesman, there are a lot of commercial companies and pro photographers whom the digital manufacturers made their mouth shut. That’s how the digital is winning the battle against the film, that how they are making their high record profits. They just don’t care about the art of photography; they hate our spirit and passion. Those things are holding them down of making even higher profits. But, there’s nothing unusual – it’s the world we are living in. </p>

<p > </p>

<p >I won’t be surprise if in the beginning of next year Nikon will announce about ceasing production of all film related equipment (2 cameras and 3 scanners). After that the entire cultural world will be duped. </p><div></div>

[greg_campbell1]

<p>Well of course you've been duped. Doh! <br>

The manufacturers spend hundreds of millions in advertising to convince the Zombie Nation how much better their pathetic lives would be if only they would by a shiny new camera. (Zombie obediently buys camera, discovers it doesn't automatically take Adams quality pics. In a stupendous leap of irrational, magical thinking, said Zombie deduces that it must all be the camera's fault! (Zombie Nation loooves to point the finger of blame.) "Good news brother! All you need is an upgrade!" <Camera company deftly distracts Zombie with MTF and MP charts.> Zombie: "Ugnh!, must upgrade!" Thus is born a new convert to the Pixel Praiser Assembly. Pass the loot, brother!)<br>

Now, who do ya blame???? The manufacturers for telling porkies, or the Zombies for stupidly lapping up the unending stream of lies? (Well, OK, blame 'em both!)</p>

<p>PS, I looove Ken Rockwell, precisely because he pisses off the types of people I don't generally like. (Hint: He clearly doesn't take himself all that seriously; some folks around here could learn something from the man.)</p>

[mjferron]

<p>Jeff I have a Epson 2400 and it gives me great B&W prints. I scan my B&W negatives with a Konica/Minolta Scan dual 4. Contrasty negs don't scan as well as thinner negs. It's easy to take a low contrast b&W negative and then adjust levels in PS for great tones. BTW developing your own B&W film at home is not that hard nor do you need a lot of equipment..</p>

[jeff_z.]

<p>Thanks Michael. If memory serves, I think the big advantage that the 2400 offered over the 2200, was greater ability for B&W. Good to know about the contrast factor.. these prints I've gotten back so far seem to have incredible contrast (in a very pleasing way), so maybe this is not the best film for trying to print my own enlargements.. I will definitely do more reading in the various forums to try to learn more.</p>

<p>Roman, I couldn't agree with you more. Also, I have the n6006, the US version of your camera. They are incredible picture taking machines, and as you point out, great bargains. I just hope that you are overly pessimistic on Nikon dropping those items.</p>

<p>Greg, I think you make some very cogent observations!</p>

<p> </p>

[Dave Luttmann]

<p>Jeff,</p>

<p>I use an Epson 3800 for color and a 7600 loaded with Carbon Pigment for B&W. My scanner for 35mm is an old Minolta Scan Dual IV. As I prefer working with Ilford HP5 and 800 and Neopan 1600 @ 1000 or 1250, resolution isn't really what I worry about too much. The biggest I do B&W 35mm is typically 14x21....and the Scan Dual IV is sufficient for that.</p>

<p>For color, the New Ektar 100 is incredible. Crops I've printed from 16x24 show very fine, tight, sharp grain....but little of it. For positives, I stick with Astia.</p>

<p>Have fun....and spread the word to help people avoid being duped!</p>

[jeff_z.]

<p>Dave, thanks for the good info. Obviously, I have a lot to learn regarding B&W, but it sure is fun! I have a pretty decent scanner, the Nikon Coolscan IV, and am pretty experienced with it and with printing color. I wouldn't want to print anything even as large as you mentioned, maybe up to 8x12, or so..</p>

<p>I don't have the time right now, but as I have the 2200, which is apparently a bit limited as far as its ability to print B&W as well as the newer models, would you recommend looking into BO (black only) printing with it? I'm only vaguely familiar with this method, but remember reading about it a while ago, I think the very good article was by Curtis Jones.</p>

<p>Yes, I've heard great things about the new Ektar, and look forward to trying it. My favorite positive film is e100g, and I've shot a fair amount, but I've heard good things about Astia, too.</p>

<p>I'll have to check into the B&W emulsions you mentioned, too. Don't worry, I definitely spread the word- it's all I can do to contain my enthusiasm for the results I get, especially with this 1970 vintage rangefinder and the T-Max 400! I wondered what you thought of this film, and your reasoning for the ones that you mentioned?</p>