 DENNIS E. GRANZOW |
resolution and resizing images If you change the resolution, say in photoshop, from 72 pixels per inch to 300 pixels per inch, and change the image size from say 2000 pixels X 1500 pixels, to 3000 pixels by 2000 pixels.....will this improve the quality and allow you to print a larger image.?
|
|
|
|
||
|
James B. Phillips |
You can always cut down the size without losing quality (much), but you can't enlarge without losing quality. Always start with the largest size you can get with your camera, and you can resize smaller as you need. Hope this helps.
|
|
|
|
||
|
DENNIS E. GRANZOW |
Then under what circumstances would you ever increase resolution and height/width(in pixels)through your editing software?
|
|
|
|
||
|
James B. Phillips |
I can't think of any situation where you would want to increase the resolution just for the fact that it would make the picture more "grainy." Perhaps for some artistic reasons, but not in general. I know that for the web, you want like 72 ppi, and for printing, you want somewhere around 240-300 dpi, but that's a generalization in itself.
|
|
|
|
||
|
Christopher A. Vedros |
Actually, some image editing programs (like PhotoShop) will do a process called interpolation when you increase the image size (number of pixels). It uses the surrounding pixels to fill in additional pixels. This can work very well, if done in moderation. Most guides that I've seen recommend increasing in 10% increments. So, for your example of increasing the long dimension of the image from 2000 pixels to 3000 pixels, you should do it 200 pixels at a time, rather than all at once. The practical application for this would be to print a larger print at a good printing resolution. When printing, the closer your resolution is to 300 dpi, the better. If your image is 2000 pixels by 3000 pixels, and you want to print an 8x12, your resolution would be 250 dpi. If you want to print a 16x24, you can leave your image as is and print at 125 dpi, or you can interpolate it up to keep your print resolution higher. A well-interpolated image printed to 16x24 at 250 to 300 dpi will probably look better than the original image printed to 16x24 at 125 dpi. Chris A. Vedros
|
|
|
|
||
|
John G. Clifford Jr |
The goal with any print should be to print at either the printer's native resolution in dots-per-inch (dpi), or a dividable fraction of that number. In order to do this, you have to know what the printer's native resolution is. It's not always 300 dpi. Let's take the Epson R300 as an example. This printer's native resolution is 720 dots per inch (dpi). Believe it or not, with good vision and your eyes pressed right up against the print, you can see the difference between 360 dpi and 720 dpi... but most people don't view prints that closely. I try to print at 360 dpi for everything under 8x10, and will go as low as 240 dpi for larger prints. Note that 240 is 2/3 of 360 (evenly divisable) and so the printer can handle the mismatch to its native dpi well. Even so, I often have to up-rez (interpolate) my image files to get the resolution I want. If you use PhotoShop's built-in bicubic interpolation routines, it has been discovered that up-rezzing in stages of 10% at a time gives better quality than up-rezzing to the final size in one step. So, if you want to double your image file size, you have to up-rez by 10% approximately 6 times. Here's the process I follow: - start with my master file Hope this helps...
|
|
|
|
||
|
Linda Buchanan |
I get my prints from MPIX and they are set at 250 dpi. Would I then set the dpi in PS to 250 when using this method to increase my image size?
|
|
|
|
||
|
John G. Clifford Jr |
Again, the 'dpi' value in and of itself is useful only to the printer driver. It does not reflect the size of the print. If you want to make a 5x7 print @ 250 dpi, then your print needs to be 1250x1750 pixels in size. That way, one image pixel = one print pixel. You'd want to resize your image in Photoshop, and since that size is smaller than most dSLR images, you'd resize smaller using bicubic sharper. If you wanted to make a 20"x30" print @ 250 dpi, your image needs to be 5000x7500 pixels in size (37.5 MP!). You are going to have to up-rez. As I explained before, up-rez repeatedly by 10% until you're close, then up-rez to the final desired size. You'll use bicubic smooth for these up-rezzes. Once you're done, set the dpi to 250 (for the printer), then sharpen using unsharp mask, and verify that you're not oversharpening by looking at the image at 50% size. Once it looks good, save it as a TIFF and send it to your printmaker.
|
|
|
|
||
|
Stephen R |
How do you find or calculate your printer's native resolution(PNR) using mfr's specs of 'optimized dpi', e.g. 5760 x 1440, that is Epson's spec for R300 and my SP1400. Also have Dell 926 AIO, max 8X10 or 8.5X11. On a very detailed moderate contrast image printed 8X the difference between D and E at 250PPI(DPI?) is for any practical purpose negligible. I do not know what the 926 res is. Anyone? Knowing that, I read it is preferable to print with the 'image dpi'(ppi?, don't understand 'image dpi')evenly divisible into the printer native res. My image is 2000x3008 pixels printed 8"x12", should I crop or something to change the 3008 to 3000? And printing at 250PPI is correct only for printer's with PNR of 500, 468.75, 406.25, 375, 312.5, etc? I know there probably are no printers with such PNR, just trying to figure out pixels, etc., that I've learned bear no resemblance to silver.
|
|
|
|
||
|
Richard Lynch |
" If you change the resolution, say in photoshop, from 72 pixels per inch to 300 pixels per inch, and change the image size from say 2000 pixels X 1500 pixels, to 3000 pixels by 2000 pixels.....will this improve the quality and allow you to print a larger image?" "Improve" may not be the right idea behind it, but you can successfully upsize images within limitations. It is probably often better to give a printer the resolution it needs (within a range) than to just give it too little and hope for the best. Upsizing is not infinite. You will not garner more detail by increasing the resolution. All you really get is Photoshop's best guess as to what should be inserted, based on the type of resampling you choose. This leads to the image looking a bit softer as you move to larger sizes. The difference in the resampling types mostly has to do with the amount of sharpening each adds. Using the resampling with the highest level of sharpening (Bicubic Sharper) is not always best if you will be upsizing the image a lot (more than 20% total), but some of the result will depend on the image and how the preset sharpening reacts to that. Stepped plans like John suggests can seem to work better on certain types of images, but not all. You can get away with more and less resolution than a manufacturer recommends, but it really depends on the application/media and expected results. Devices will need to use different resolution based on their processes (inkjets might use less than halftone processes, for example). The way the printer uses information varies even within the same printer type. One size does not fit all. dpi is Dots Per Inch -- a printing term, really it is the PNR as you describe it ppi is Pixels Per Inch -- a digital image resolution designation for the preferred use in the file (it may be ignored) Printer resolution is often quoted by manufacturers in confusing ways. For example, a 5760x1440 resolution printer that has 8 inks is really similar to a 720x720 4 ink printer as far as number of dots per ink. The 'higher resolution' printer accomplishes its higher side by adding a half or in this case quarter step between ink rows (5760/4 = 1440). If you divide the total number of dots by the number of inks, you get the real absolute smallest dot: 1440/8 = 180 This is why some exclaim that 180 is the low end for resolution in an inkjet. Calculations differ, but unless I have another reason, I use the following for inkjets, staying close to the upper end as a personal preference: [1 to 1.4] x [dpi/inks] That is, for a 720ppi printer with 4 inks: [1 to 1.4] x [720/4] Somewhat higher than that (10-20%) is 'safe', but recommendations vary. My course From Monitor to Print: Photoshop Color Workflow covers all this and more, getting you up to speed on not just rendering your images at the right size and resolution, but with the color you expect as well. I hope that helps! Richard Lynch
|
|
|
|
||
|
Stephen R |
Richard, 1) So, on a 5760 X 1440 dpi 6-ink printer you're saying my 'PNR' is 240DPI(1440/6); why isn't it 960(5760/6)? 2) And does it really matter? In your eg you take that short-side PNR and multiply by 1 to 1.4 = 180 to 234. Are you saying to size the image to anywhere between 180-234 PPI to print? If so, this contradicts what John said earlier about making the image PPI evenly divisible into the PNR. 3) And your PNR calculation method conflicts with what John said about his R300; it has, just like my SP1400, 5760X1440 with 6 inks; by your method that PNR is 240; he said his PNR is 720??? These apparent contradictions have me confused about what is real and what is opinion.
|
|
|
|
||
|
Richard Lynch |
Stephen, I don't mean to contradict anyone. I believe I left enough play in my response that it was general while having some play for different machines and manufacture (e.g., "but recommendations vary"). 1. The 5760 is just a quarter step in the movement of a row. The lower number (1440) is the actual finite, maximum resolution of the printer in that it is the smallest dot it can make. Thaty the printer has the ability to overlay dots does not make the dot size potentially smaller, it just offers the opportunity to overlay them. 2) My prepress background has some bearing here in that calculations are usually made for PPI matching DPI plus some overlap for interpolation/blending. A 1 to 1 ratio is optimistic, but some manufacturers have claimed it to be valid (e.g., factoring bleed). I frankly don't think that there is any contradiction depending on how you choose to look at the numbers. I don't, however, think you can choose a range as a one-size-fits-all...What works for one printer type and perhaps even manufacturer is not exactly the same as another. I think John was speaking from his preference and experience. He chooses to go 240 minimum to 360...and these are valid. I was just pointing to a possibility on the low end of the range and a considered calculation. 360, I believe, is more than an inkjet will generally need, but it is not so obscenely high that I can say "yikes, that will take a century to process!" It will likely be more than needed but with the speed of computers what it is, the time in processing will be negligable. 3) The difference noted may be in understanding, or in the way most manufacturers denote their resolution. Manufacturers count all the colors because it is more impressive. To one extent, you can see that if the finest dot the printer can put down is 240 dpi, that really doesn't increase because the printer can use more colors or overlay more rows. DPI is the measure of the absolute finest particle of ink the printer can put down, and the measure I like to consider is much like the way dpi is used in offset/halftone printing. The comparison is a little like apples and oranges because inkjet printing is far more like stochastic printing than printing halftones. However, I think that taking the smallest potential particle is the true dpi. One thing I cannot reconcile from John's post is that a 6 color 1440 printer is 720 PNR, as it doesn't follow either of the standards. Perhaps John can fill in that detail. My interest is not to prove anyone wrong or find contradictions. Usually what is happening is there are different means of interpreting -- sometimes used as marketing ploys and sometimes that merge with other facts. I think John is adept in his understanding, and that more than one explanation does not necessarily denote contradiction. Richard
|
|
|
|
||
|
Stephen R |
Richard, Maybe 'contradiction' was too severe; I'm not trying to belittle anyone, tell anyone here they don't know what they're talking about, prove anyone wrong or find contradictions. I'm trying to learn this digital photography so I can be as proficient as I am in film. I've been shooting for over 25 years and jumped into dig 2 years ago. My current 'methods' are working OK(without knowing all this dig. theory), I just always like to know 'WHY' something is so. I just think(thought) these were apparent contradictions or perhaps differences in terminology, etc. I do know Math, and thus: On #2) Still think there is a difference in the 2 methods discussed; John says size image so PPI is a dividable fraction of printer's PNR, although in his eg he says 240 is 2/3 of 360; technically in math-speak 240 is NOT a dividable fraction of 360 because 360/240=1.5. Dividable fractions mean the result is a whole number. So any fraction, e.g. 5/12, could be applicable as 360/150=2.4 also not a whole number. His example(not his divid. fract. statement) appears to concurr with your example of 180-234; as I understand you, you're saying adjust the PPI of image ANYWHERE betw 180 and 234? e.g 217 or 189? Again, I question does the ACTUAL value really matter as long as it's within some acceptable range? 1) Is that 240DPI correct as my printers native res. using your method? The idea of this PNR being calculated using # of cartridges in printer has me flummoxed; My Dell has only 2 cartridges, one black, one color or photo. Of course that 2nd cartr has multiple colors within. By your eg if Dell is 1440 on shortside, it has 720DPI PNR -- so the less cartridges the higher the PNR!!!??? I've played with all these numbers and note using 1440 as printer res(PNR) makes John's figures and statements work fine, i.e., 1440/360=4, 1440/240=6, Whole Numbers. Why not use 1440 as the actual printer res? But again, you say size anywhere betw 180-234, so if I pick(extreme # for my EG)217, print will be OK because it's within that range, but 217 is not any kind of fraction of 1440 or 240, whichever is used as PNR. 3) If DPI is the measure of finest particle the printer will lay down, 1.5 picoliters on 1400, what does the # of cartridges have to do with this? Isn't Epson's 1440DPI value arrived at by taking the 1.5pl into account? If so, I'm leaning towards using 1440 as my printer res. and then sizing image so that my PPI divides evenly into it, i.e., Whole number. IF that's important as John states -- I see logic in that, but then using that 217 figure I picked from your range appears to make all this discussion a moot point. Maybe I'm over-dissecting this whole issue, I don't know. I've done OK with all this in a practical sense, my prints are excellent without doing any of these calcs, but knowing technical matters better from a theory standpoint can only improve things I believe. So far I've always set the PPI of image to 300 or less, down to approx 180 as I heard/read this is a good range for printing, and it works. Again, I'm never trying to question anyone's skill, technique, knowledge, etc. to try to prove them wrong or attempt a put-down. There are individual preferences, opinions, etc. and then there are facts. I like to hear both from respondents, but want to determine what is factual theory, because I have preferences that seem to work, but don't always know why or if my idea about such is even correct theoretically. Of course I could run my own tests to prove/disprove my ideas about printing, res., etc., but I'd be using lots of paper & ink for such; I've already used a fair amount running smaller tests in the past years. So, I thought it's simpler to ask in a forum from those already familiar with these concepts. Man, if pixels were capable of moving around the image like silver, we could just, say, print at 8X, 11X, etc. and be done with it! Thank you
|
|
|
|
||
|
Richard Lynch |
Stephen, I didn't think you were looking to stir anything up, I just didn't want to appear to. John participates quite a bit here on the BP forums and I don't want to appear to be overriding what he is saying--and I think there are ways to meld the ideas. I have found his contributions to be valuable and positive, and he is a good resource. 2) You mention that there 'may be a difference in methods' and that is more likely than not. I tried to suggest a background or resource used may point to a different calculation. The 'science' of the calculations is a little blurred, as there is a difference between equipment and philosophy (in a similar way people might make arguements for or against CMOS and CCDs, or Bayer and Foveon sensors). I am not sure about John's method for derivation, and perhaps he has an explanation. My derivation has to do with attempting to calculate real, finite measure of the possible smallest printer element (dot) so that can be used to determine the resolution you need in the image. Usually (and that is an indefinite) manufacturers provide total dots as a value indicative of: dots x inks (not cartridges) So if there are 6 inks and 1440 dots, the printer is actually capable of 1440/6, or 240 dpi That *usually* holds, but marketing people sometimes help the manufacturers to make confusing and opposing data filed as the same spec type. This is why the manufacturer may have different recommendations. Mine go by industry trend (which can change and I have generalized). As some manufacturers claim you only need a 1 to 1 ratio of image pixels to printer dots for printing (this may be an exaggeration, in halftoning the ratio is more like 1.55-2), and others claim a good approximation is 1.4, my calculation of 1 to 1.4 pixels per dot may be somewhat conservative considering...yet it has played out well in practice. The idea is that you want at least as many pixels as dots, but likely more so there is a good blending of information to create consistent tones. The actual ppi value has little importance within the range -- as long as it is within the range. You just don't want too much (over-processing) or too little (pixelization). 1) use number of inks, not number of cartridges in making this calculation -- at least my way. As I learned it for printing and pre-press, this makes most sense to me. We'd often just print in black. If you do that, your printer would suddenly become a 240 dpi printer, and, well, that doesn't fit with the idea that printer dots are the finest element you can print. When you add colors to a multi-million dollar press, it doesn't leap from 3000dpi to 6000, 9000 and 12000 (for CMYK) as it would by this inkjet "standard"...it is still 3000dpi. It just uses more color. Part of the confusion here is willfully caused by marketing gurus who don't care if you understand it, they just want to sell it. The smallest printer element is not multiplied by the colors you apply...it is the single smallest dot you can make in any one color -- at least from a pre-press standpoint. 3) I admit to not having the foggiest what a 'picoliter' was except that I guess it is the volume of ink in a dot. My guess is that most printers are similar in this reguard in resolution, as most factory presses are similar in resolution (depending on type/use). More expensive printers might have fractionally smaller picoliter measures -- but I am guessing. The dissection is getting too small. We might also have to get into chemical breakdown of the ink to define drying time and volume reduction...meanwhile the manufacturer claims a dpi. The latter seems suited to our purpose. That numbers fit conveniently and make pleasing calculations doesn't change the matter. Along that line, I am not clear on the importance of fitting your ppi evenly into a printer resolution and what advantage that provides. My take: if you have too little information, your image pixelates; if you have too much, your pixels need to be crammed into a dot, and the printer will end up making calculations and decimate (interpolation term for remove) or merge information in the file. It will do that whether you use 240 or 241 ppi, and both will print fine. In fact you can print and make prints nicely anywhere in the range -- test it out. Likely you will see little or no difference except under a microscope...and then you will end up counting picoliter residue. In all, part of the key to getting what you need in print is providing the printer at least the minimal amount of information it needs to satisfy its resolution. If you feed it too little, you may get noticeable pixelation, banding, and other issues. If you feed it too much, you make it work too hard for no better result. The most difficult part is defining the range, because industry 'standards' have become such a mish-mosh of terms and strategy to gain a marketing edge. Richard
|
|
|
|
||
|
Richard Lynch |
By the way, you say: "Man, if pixels were capable of moving around the image like silver, we could just, say, print at 8X, 11X, etc. and be done with it!" There is a finite resolution to film as well. Pixels can be treated in nearly the same way. Depending on the interpolation properties of the process, it will work the same way. As an aside: Frankly I never print at home. I send my images out to a service where they use printers I could never own because they are so expensive, and my cost actually goes down in the process (these printers are about 100 to 1000 times the cost of a home inkjet). By establishing confidence in what I see on screen, defining a logical workflow (color management settings, image resolution/sizing, correction techniques), and properly submitting images I get consistent results with less cost in a greater range of color. The From Monitor to Print: Photoshop Color Workflow covers all you need.
|
|
|
|
||
|
John G. Clifford Jr |
This has turned into a very interesting thread. I'm sure that Richard has much more experience with high-end printers than I do, and ditto for his post-production experience. I think the reason that Richard's and my recommendations differ is that he has more faith in the technology (or the developers of that technology) than I do. That faith may well be justified if one works with high-end printers a lot, but I work with the printers where the ink cartridges can cost more than the printers! (Epson R300, 2200.) The reason that I like to get a base number, e.g., 720 ppi, and then ensure that my number is an even multiple of that (180, 240, 360), is because, as a software engineer, I understand how I am greatly easing the printer driver's job of interpolating between the dpi of the image and the printer's native dpi. I don't want the printer driver to have to guess; I want it to do what _I_ want it to do, not what some software developer thinks it should do. Re resolution being related/dependent to the # of inks... this may be true for some printer designs, but I know that the Epson family of printers can print at their native 720 dpi resolution with one color, e.g., black and white. The printer is smart enough to know about the physical design of the printhead and how to move it to get the full resolution. I have done some informal testing with odd, arbitrary dpi settings, and have found that there is a perceptible quality difference. Anyone with an inkjet can try this at home; try printing one image with 240 dpi and then the same image at the same print size, but resize/re-rez the image to have 255 dpi. You'd think the latter would be sharper... but it will also have visible artifacts. Lines may not be as sharp, etc. I've also up-rezzed a few images to 720 dpi and printed them at that resolution, and they ARE sharper than a 360 dpi print... if you stick your nose up to the print. You'll never see the difference from even a foot away. (I was thinking about printing a huge panorama at this resolution but found that it really wouldn't make a difference.) Really, printer drivers are RIPs (raster image processors). They have to turn the bitmap in the image file (and every image file is, at heart, a bitmap) into a bitmap that the printer can render with its fixed native ppi onto a piece of paper. The printer driver WILL scale the image. My goal is to give the driver the easiest scaling job possible by letting it use integer math so that it can't screw it up. You know, it really doesn't cost a lot to make a 4x6 print on a home inkjet printer. Experiment!
|
|
|
|
||
|
Stephen R |
As I understand it & tend to agree this may be best, John says to size the ppi of the to-be-printed image so it it a RATIO with the PNR, e.g. PNR is 720dpi, make image ppi 240ppi, 3:1 ratio. This results in less of that odd number 'interpolation' done by the printer. Logical to me. At say, 250ppi, there may be more evaluative, destructive?, decimative interpolation being done by printer than at 240ppi. As I understand Richard, that intepolation is going on anyway, getting 240ppi to 720dpi so it doesn't matter if printer interpolates an even ratio(720/240) or not(720/250). Questions: John, I looked up the R300 Epson and it has the same 1440X5760dpi 'optimized res.' as my SP1400. How did you arrive at 720dpi PNR for your R300? Why don't you use 1440 as PNR? BTW, Richard, I know film has limitations too, but the stationary pixels of digital necessitate these considerations when enlarging(I believe). Also, perhaps that's part of the reason dig doesn't have the exposure latitude of even Kodachrome 64 or 25, my John, I can't typically see differences printing 4X6's even under magnification. How do you?
|
|
|
|
||
|
Richard Lynch |
Stephen, I think the idea of the ratio is summed up fairly well from my perspective in the first paragraph, but perhaps goes a step further (or half step ;-). I cannot tell you the type of interpolation that a given printer uses in calculating pixel use (or the computer doing the processing), but I know interpolation in Photoshop quite well. I would assume that for most any circumstance the result of interpolation in the printer (using Photoshop terms) is the equivalent of Bilinear interpolation rather than Nearest Neighbor...The difference between these being Bilinear is a straight average of pixels in a dot area by percentage: [I know this exact relationship will not happen, I'm keeping it simple] pixel A is 20% in the dot area and pixel B 100% in the dot area, and the result is 16.7% the color/tone of pixel A and 83.3% the color/tone of pixel B. Nearest Neighbor would be more like pixel A is 20% in the dot area and pixel B 100% in the dot area, and the result is the color/tone of pixel B (it chooses one or the other). The latter would likely require nearly as much calculating power as the former, but almost necessary lends to getting a bad result (jagged edges or pixelation -- no matter how slight). Whether the printer dot has 1.2 or 1.55 pixels in it (1.55 is a number often used as a suggested minimum for halftone printing), some calculation needs to occur (as you summarize) whether the decimation is an even or odd percentage it is still a calculation and a new result. In my estimation it would be difficult to see, if not impossible, even with a microscope...and that one is notably better than another in fractional difference is unlikely. The only time I find that calculating pixels to a ratio is important is in upsizing computer screen shots. They are 72 ppi, and rendered in a blocky fashion from the outset. To size them I upsize to 288ppi (4 times the original exactly) using Nearest Neighbor so that each pixel turns neatly into 4 -- with no blurring or fractional interpolation. I would never do the same thing with an image (Nearest neighbor interpolation) as if you are going to upsize, you generally want to maintain blending between pixels, and fractional interpolation might reduce rough spots (again, pixelation and gradients). In other words, being too neat about it may not do you favors. I'd be curious to hear more from John about the calculation as he understands it. I would hope a printer would not render an image with jaggedness and pixelation in an uneven increment/ratio of pixels to PNR. While I am limited in experience with inkjets, I've not seen this behavior and wonder if there isn't some other means of control if pixelation is a result (e.g., printing at the highest quality a printer is capable of). A manufacturer requiring people printing at home to follow a ratio between PNR and ppi would likely not be doing a good thing, and they'd best mention it in their manuals. Exposure, in some sense, is limited just as much by the output as the capture. I don't know of any output that is more than 8-bit, and that is a limitation born of the computer world. Higher bit processing and other enhancements may lead in another direction (digital is still relatively new), but printing processing and devices need to make the leap (Photoshop already stores 32 bit images with HDR). No argument about latitude, but often better latitude can be simulated by better correction of images, which I teach to some extent in all of my courses. As far as the enlarging, I guess my point was you have the same limitations with enlarging film--you eventually reach the finest point in the medium. The difference is digital can be resampled -- of course not to bring in more actual detail -- but in a flux. Beyond that, it will blur, as film might. I guess I see more similarity than difference, but it certainly depends on how you look at it. John said: "I've also up-rezzed a few images to 720 dpi and printed them at that resolution, and they ARE sharper than a 360 dpi print..." I would think that is a result of the Bicubic interpolation (in any of its forms) which adds sharpening. The result will be different and appear sharper but it is not necessarily a result of the ppi as much as the process of interpolation that gets it there. I think using Bilinear interpolation for the test may render nearly exactly the same results in the 360 and 720 ppi print as it will not sharpen the result. I hope that helps! Richard
|
|
|
|
||
|
DENNIS E. GRANZOW |
I finally went back to read all of this again to see if I could understand it. I'm finding that it is way over my head no mater how many times I read it. However, I really appreciate the time you have all spent in trying to help me. If I could ask one more thing please...Let me give you a brief simple scenario that I think can be answered in one brief paragraph that is not too technical. Let's talk in "generally speaking terms" and see if it helps. Here's my situation....I used a Canon Rebel XT to shoot a picture at the highest quality setting (3456x2304). I uploaded the shot to my computer which is now a JPEG file (size 3456x2304 pixels, 48x32 inches with a resolution of 72 PPI) Question...what adjustments, if any, would I have to make in order to make a print 48"x32"? Let's say I crop this picture, and it is now (2084x1441 pixels, 30x20 inches, with a resolution of 72 PPI)What adjustments would I make to print it to a size of 48x32 inches. Is it possible to tell me what to do without telling me why at this point. Thanks again for all your help.
|
|
|
|
||
|
John G. Clifford Jr |
Dennis, this is 'easy', and serves as a good example for illustrative purposes. Your image is 2084x1441 pixels. The ppi doesn't really matter, as that is a function of the display/rendering type (monitor, printer) and not the image per se. You want a 48x32 image (landscape aspect ratio, 3:2). In order to get an acceptable print at this size you'd want at least 180 dpi, and 240 dpi would be even better. Let's first figure out how big the image needs to be to get a 48"x32" print @ 180 dpi. Well, 48" times 180 pixels/inch = 8640 pixels... your image needs to be up-rezzed from its initial horizontal width of 2084 pixels until the horizontal width is 8640 pixels. That is more than a 400% enlargement, so chances are this isn't going to be a great looking image from close-up. I find that, with Bayer sensor-equipped digital cameras, the image quality starts to fall dramatically once an image is enlarged more than 200%... yet another reason for the good advice of cropping in the viewfinder (using a longer focal length lens). Anyway, sticking to the image we have, I'd up-rez it, by 10%, using bicubic smoother, until the image was very close to my desired size, 8640 pixels, and then the last up-rezzing would be to the final size. Because I'd do this while maintaining proportions, the 'short' axis of the image would increase proportionally so that it was the correct sizw as well. I'd end up with an image that was 8640 x 5760 pixels in size. Once I had that image, I'd use unsharp mask and sharpen judiciously, judging the image by viewing it at 50% of actual size, and tinkering with the sharpening (undoing and redoing it with altered settings) until I was happy with the image. I'd save the final image as a TIFF file, and then send or take it to my printers, telling them what size print I wanted at what dpi (48" x 32" @ 180 dpi). There you go!
|
|
|
|
||
|
Stephen R |
Hi John, As my mailbox alerted me to this thread today and I saw you responded, thought I'd ask; How DID you arrive at that 720DPI PNR for your Epson R300? Why not 1440DPI as published by mfr? Thanks
|
|
|
|
||
|
DENNIS E. GRANZOW |
Stephen...thanks so much for your patience, I really appreciate it. I will follow your instructions and get it done. I now know how to do it. But, I'm still a little confused about the relationship between dots per inch and pixels per inch. I thought DPI was a printer term, and PPI was an image term for resolution. How do the two relate?
|
|
|
|
||
|
John G. Clifford Jr |
Steven, read my comments in this thread (http://www.betterphoto.com/forms/QnAdetail.php?threadID=19495) to get the long version as to why I believe 720 dpi is a valid figure to use for the R300's native resolution. The short version: testing has shown that desktop inkjet photoprinters produce the best results at 2x the hardware's 'native' DPI.
|
|
|
|
||
|
John G. Clifford Jr |
Dennis, DPI (dots per inch) and PPI (pixels per inch) are both used as units of resolution. With an image file, the number of pixels per inch depends on the hardware characteristics of the display device. If you're using a typical LCD monitor to display your image, then the PPI figure is usually around 90, and the size of the image as rendered will be the number of pixels divided by the PPI. With a printer, DPI is the size of the ink 'splat' used to display a pixel. For instance, if I print a 2700x1800 image at 300 DPI on a printer, then I will get a 6" x 9" landscape print. So, you use the DPI setting to 'scale' the image with a fixed pixel count to the print size that you want, e.g., using 150 DPI with the same file would give me a 12" x 18" landscape print. The reason we care about DPI as well as PPI is because we get the best print quality when we take the printer's native resolution, in DPI, into account when sizing the image for printing. A quick way to verify this is to print an image at evenly divisible multiples of the printer's resolution, e.g., use 360, 240, and 180 dpi for a small section of a print that is sized to 720 x 720 pixels. Then, try non-evenly-divisible multiples, e.g., 325, 255, 170 dpi. Compare the results on your printer.
|
|
|
|
||
|
Stephen R |
Dennis, In theory, there is no relationship betw. DPI and PPI. You are correct that DPI is strictly a printer spec. However, PPI is a RESULTANT of the number of pixels vs. the print size, not technically an 'image term for resolution' -- more like an 'image vs. print size term for sharpness'. An image, per se, does not have a PPI value(I know the editor shows one, but that's only because it also shows an intended print size, thus you see a "PPI" value). I ignore these 2 figures in the editor; before I crop, resize or print I decide what print size I want and what PPI value I want in finished print; then, is it doable and how, crop, resize w/o resampling, resize w/resampling; then I edit image size to accomplish that; then, I go to print window, enter one dimension in print size(the other one will automatically be the correct value based on how I adjusted the image size to accomplish desired PPI in both axes), and print. Now we come to the 'in theory' part abt. DPI vs PPI relationship. Until I read John's msg about having a RATIO between 'PNR' and PPI, I never gave it much thought and didn't see a need to establish ANY kind of relationship between the 2. I've even read in Pop Photo by Debbie a few months ago to 'let the pixels fall where they may' -- no relationship at all -- as long as your resultant PPI is within an 'acceptable' range(that ubiquitous 300 or so). But I've run some tests based on John's ideas and decided it can and often does make a difference in finished print, BUT sometimes not. I use a 'PNR' value of 1440DPI and size/print so that the resultant PPI is one of these values -- 720, 480, 360, 288, 240, 206, 180, 160, 144, etc, i.e, ratios of 2:1, 3:1, 4:1, 5:1, 6:1, 7:1, 8:1, 9:1, 10:1. However, John uses a PNR of 720DPI on his Epson R300, and it has the exact same spec as my Epson SP1400, 1440DPI on the short side. I wish to know how or why he arrived at that figure. Using 720 limits the number of DPI/PPI ratios to half of what I've stated -- 360, 240, 180, 144, etc. The tests I've run show me there is an advantage to his method in certain cases, and I have concrete evidence of this(using 1440PNR not 720). I don't want to get too lengthy here on those cases where it was superior, unless asked. And my evidence used one PPI value not available using 720 as PNR, i.e, 206PPI(and the weird part here is that the true ratio using 1440PNR at 7:1 is 205.714..., not exactly 206!) But it shows an advantage over the same print at 255 on my Epson. So again, John, why 720?
|
|
|
|
||
|
Stephen R |
Sorry guys, My last response was after I saw Dennis' last msg, but b/4 either of John's last 2. I haven't yet studied, only glossed over, the thread John advised me on, '#19495', or the Clark posting. But from what I did see, I must study this further(verry interesting). And I will have to test that 206PPI vs 240 and 180, also 288 vs. 360 and 240. I never tested the 1440 vs 720 issue. From what you're saying, I should see an even bigger improvement between a 240 vs. 255 print than was evident betw. 206 vs. 255? If not, I'll keep using the 1440 figure, as that 206 vs. 255 test was solid proof, at least IMHO, of the value of your method. Thanks, John.
|
|
|
|
||
|
Richard Lynch |
The numbers of print dpi vs. pixel ppi become nearly incalculable in modern inkjets because of the plethora of configurations -- and also the inadequacy of manufacturers to accurately report the resolution of devices in terms that are consistent and reliable. Does one pay attention to the number of print heads? The number of nozzles? The size of the droplets (picoliters)? How does this mesh with image resolution and determination of PPI? A printer that claims to have 1200 dpi is measuring what exactly? Is it the number of dots total per ink per inch, or the combined possibility of the number of inks per head? The number of nozzles per head per ink? In actuality the newer measures may have become more important than the older concerns of dpi. Regretfully none of this is standard. The HPZ3100 for example claims 1200dpi with 12 inks but is not using one head per ink -- two inks per head. The droplet size of 4-6 picoliters is not the smallest by any means in the industry...and 2112 nozzles not nearly the most per head...So why does this printer cost $5000 or more in the 44" version? It may not be as much for the quality of the printing as the availability of printing 44" x X" at home on roll paper. But then there is the other matter of how large the print head, how dense the nozzles...how do the images separate...what color management settings do you use that it interprets...All this affects the rendering, and the horribly complex litany of parameters and measures cannot in any way be summed up by a generalized 'notion' of what dpi you should use to get the most from any particular configuration and technology. Then there is the small matter of what the bare-naked human eye can discern as difference. Unless your eyes are very special, they will see a blend of droplets rather than the droplets themselves. Printing with obscenely high ppi will perhaps serve (beside slowing down printing and actively encouraging unnecessary interpolation) to make a difference in the array of vapor droplets when they are nozzled out to paper...but in reality there is a point where the human eye will no longer discern real difference in detail. Whether the difference in array actually matters on a microscopic level likely becomes academic. Do I think there will be a difference in printing a 720 ppi image over printing one at 360 ppi? Yes. And you can show it. However, I also don't think rightly this will matter in detail to the naked eye (leave your loupe at home), and I wonder if it is really 'accurate'. Resolutions of 650-1000 ppi are often reserved for printing to machines with such fine resolution as to render slides, negatives and transparencies...all of which are created with the intent to be enlarged. Is an image really required to have such fine detail that you can take out a loupe and inspect differences and be satisfied that the picoliters do not fall the same way? Can it actually be seen? More...is it actually rendered? I am not sure that the latter is even true. I am pretty sure that if you have to get out your microscope to make the comparison that it is moot. More if you have increased the resolution of the image by 2 fold, you are comparing apples to oranges. Upsizing an image from 360ppi to 720ppi is not adding more detail -- it is interpolating and making up image information that did not exist prior to upsizing -- an educated guess based on a mathematical calculation, done strictly by a device that has no capacity to see or judge content. It will be a different image after interpolation, but it will arguably also not have more detail or more accuracy than the original. The calculations I see here are trying to make an absolute determination where I do not think it exists. There is a range of acceptable definition that one can hope to achieve at finished size, and a realistic notion that the image as you shot it is the image with no more detail ever to be found no matter how you upsize, sharpen, magnify, or process. Interpolation is creating faux information. Taking your 3000x2100 pixel image and upsizing to print 3429x2400 to print an 8x10 (with some cropping) @300ppi is likely harmless, but jacking up the ppi so that you interpolate fully more than half the information in the image...I am not sure that is responsible to the original image detail, or necessary for what the eye can perceive or how it perceives it. High-end printers that render graphics in the print industry are 3600 to 5000 dpi, and the requirements there are usually never for more than 305ppi images. Granted LPI, and halftone printing are a somewhat different animal than injet printing, but the fact is the eye has limitations, and interpolated information is not adding any more detail. Most are satisfied to look at a screen or monitor that can only render 96 dpi...and even that will look reasonably sharp by the fool eye. The resolutions suggested here are not in line with sight or technology. Richard Lynch
|
|
|
|
||
|
John G. Clifford Jr |
Richard is right... interpolation is adding pixels, not detail. That's why I use Foveon sensor-equipped Sigma dSLRs. All other dSLRs (Canons, Nikons, etc.) use interpolation to create their images. The apparent high resolution images we see from these cameras are a testament to how good demosaicing algorithms have become... at convincingly guessing what the sensor missed. Because other, Bayer sensor-equipped cameras already use interpolation as a key part of the image file creation, the ability to further enlarge images is greatly reduced as the signal-to-noise ratio decreases (the image quickly becomes more 'noise', or added pixels, than original captured data). Richard is right again, when he says "the eye has limitations." Much successful photography involves tricking the eye, e.g., the use of filters to simulate lighting that never really existed when the photo was taken, or post-processing to subtly brighten the subject and darken other areas of the frame to make the subject 'pop out' at us subconsciously. Similarly, judicious interpolation fools the eye into thinking there's much more detail than was actually captured. For instance, grain can add to a sense of detail. What I try to prevent when it comes to printing is aliasing, the 'stair-stepping' of lines that is jarringly visible to our eyes. We aren't used to seeing stair-stepping in real life; when it or other artifacts such as mosaicing appears in an image our ability to view the photograph as an image is diminished and instead we see the physical photo. Sigma/Foveon plug: Look at the 'Green Heron' image in my gallery. This is a full-size crop (1 pixel on the image equals 1 pixel on the sensor) taken from the center of the original image. Move your head back 18 inches (50cm) from the screen, and ask yourself if the image is acceptable at the size displayed. In my opinion it is. Now, realize that even my cheap Epson R300 printer has much better resolution than the typical computer monitor, and realize also that I could easily enlarge this image by a factor of 2 (twice the size) without NOTICEABLE artifacts... and then realize I could print an outstanding 10" x 15" print that would knock your socks off from a viewing distance of 12 inches, and a 16" x 24" print which would look great from 2' out. That's why I prefer the Sigma, even over cameras like the 5D. BTW, I saw a local photographer's print, taken by a Canon 20D 8.2 MP dSLR, that was a center crop of an image of a rockchuck (western woodchuck) sitting on the edge of his burrow. The image was printed at 16" x 20", and while it looked decent from about 5' away, it looked HORRIBLE from 2' or closer. Why? Because the camera had already interpolated the image, and through the cropping and printing process the photographer interpolated it further... to the point where perhaps less than 20% of the pixels were what was originally captured.
|
|
|
|
||
|
Richard Lynch |
Interesting that John and I use the same camera (Sigma SD14)...My choice was also do to concern for sharpness and getting the best capture. Richard Lynch
|
|
|
|
||
|
This old forum is now archived. Use improved Forum here
Report this Thread |
||