Posted by Steve Balke on June 20, 2009 at 11:41 PM
Considering how expensive and heavy are telephoto lenses, the subject of cropping begins to look pretty important. I’ve always just cropped by eye and left it at that. However, now I’m becoming more serious about photography and thinking about buying a faster telephoto lens. So, it looks very worthwhile to try to get an estimate that will help me make an accurate decision. I searched the forums and the Internet to try to obtain some quantitative guidelines. I’d like to know how aggressively I can crop and the consequences regarding printed image quality. I realize that there are quite a few variables. However, there is some information. I invite any of you who are interested in this topic to please let me know what you think of the following points (and any other comments too of course!):
i. A range of 200 to 300 pixels per inch looks like a good range for prints. So, if I want to print a 4x6 inch print at 300 pixels per inch I should try for a cropped area containing 4x300x6x300=2.16 megapixels. (That looks good: my trusty Canon provides 12.2 megapixels). Of course, 4x6 isn’t all that large a print either! Is there a popular print size in nature photography? Are 300 pixels per inch enough for nature photographs based upon your experience?
ii. The question of whether or not the subject will fit into the minimum cropped area is more difficult to answer. Lens magnification seems a bit confusing to me. Dividing the focal length by 50 should give the magnification (8x for the maximum magnification of my 400 mm telephoto). However, the specification for the lens states a maximum magnification factor of 0.2X at 400mm. I understand that this is the ratio of the height (or width) of the subject on the camera sensor to the actual height (or width) of the subject . I know the size of the sensor from the camera specs. So, I could then assume a size for the subject and figure out the area of the sensor that will be occupied by the subject. However, that number will be independent of distance to the subject (which does not make sense!). I’m missing something! Any suggestions?
Here are the main references that I found (Know any more good references?]
http://flyingsamphoto.com/digital-photo-printing-megapixels-200.shtml
http://flyingsamphoto.com/digital-photo-guide-art-how-many-pixels.shtml
http://flyingsamphoto.com/digital-photo-guide-article-006.shtml
http://www.the-digital-picture.com/Canon-Lenses/Canon-Lens-Magnification-Value.aspx
The number of pixels in a 4x6 will need to be higher because you are typically looking at them at a distance of 18" or so. While 300 p looks nice, when you get to 24x36, which has a viewing distance of 3'-5' even 200 p or less is all that is needed. Upsizing algorithms in Photoshop or standalone print programs like Qimage will ensure that your image looks good at the larger printed size. Also, programs like Genuine Fractals will upsize an image with minimal loss and maximum quality. Lastly, while camera megapixels are an important factor, lens contrast and resolution, pixel pitch, sharpening (either in camera if you shoot jpg, or in postprocessing if you shoot raw), your tripod and head etc will all determine the acuity of the final image.
Some general guidelines - start by getting the sharpest lens you can afford. More expensive telephotos are generally sharper than their less expensive brethren, and they are also faster, so you can shoot at lower ISO and/or faster shutter speeds. Make sure your camera support system is not generating vibrations which will cause motion blur, especially when the mirror slaps up and down, and establish a workflow in your editing program that preserves as much of the original image quality as possible.
For the record, I have seen some exceptional images printed to 24x36 that were shot with a Nikon D70 (6.1mp) and 16x20 from a D2H (4.1mp). There is a software product called DXO Objects Pro that in addition to doing some "magical" adjustments with respect to lens optical corrections, it also manipulates pixels to maximize image quality, especially for large print sizes.
Posted by Howard Cheek on June 21, 2009 at 11:40 AM
I use Genuine Fractals for resampling. When I crop, my cut-off is a 17 MB TIFF file size @ 8 bits.
(Generally around 3000x2000)
The image needs to have very good original sharpness, and ISO quality for the best results.
Of course, you never sharpen the image you're enlarging/resampling.
I've resampled images using GF up to 4x6 feet in size using this method, with excellent results.
Hope this helps a bit,
Howard
Om (ii), I think you are confusing two different things. Dividing the focal length by 50 is the magnification of the subject relative to a 50 mm lens.
In your example, you use a 400 mm lens. Pick a subject that will be 1mm in the frame with a 50 mm lens. Keeping the camera in the same spot, now switch the lens to a 400 mm lens. That same subjuct will now be 8 mm in the frame - the 8x.
Now, let's get back to the lens specifications. Take that same 400 mm lens and suppose that it says maximum magnification of 0.2. What this means is that if you focus that lens at its minimum focusing distance, a object that is 5 inches long in real life will appear to be 5x0.2 = 1 inch long back to the recording medium.
A cropped sensor does not matter for the second case of lens specifications. Here's why: there is a cone of light that is in perfect focus on the plane of the sensor (or film). If you could see that light, the 5 inch long object would still appear to be 1" long, or 0.2 magnification. Putting a full frame sensor back there would record all of that one inch while a cropped sensor would record maybe 0.6 " of it. Either way, it is still 0.2 magnification, just more or less of it gets recorded.
Here is where I begin to go beyond what you originally asked in (ii):
Some confusion comes in in that last paragraph with cropped sensor. You'll see that the focal length is 1.6 times as long. So shouldn't that mean that the maximum magnification is 1.6 x 0.2 = 0.32? No, that 0.2 means what is recorded on the sensor.
If you print to 8 times the sensor size, for a full frame sensor or a cropped sensor, you wind up with the same amount of magnification. However, if you print to the same print size, say an 8x10, from both a cropped sensor and a full frame sensor, the cropped sensor would show more magnification. (Remember: to get to an 8x10 from a cropped sensor, you need to enlarge the area even more than with a full framed sensor to get to the same size, since you are starting with a smaller recording medium). In this case, you would have effectively a 0.32 magnification, not a 0.2 magnification.
Think of the 0.2 as lens/recording magnification, the 0.32 as a *effective* printing magnification.
I am still trying to figure out how to respond to (i). As you said, there are a lot of variables to deal with, some of which are at odds with each other. My best suggestion is to make some prints at various croppings and sizes and see what works for you. What I find acceptable may be garbage to you.
Dan is exactly right. Magnification is a function of focal length and subject distance. It is not affected by sensor size. The sensor size will only determine what portion of the projected image gets captured. In practical terms it determines the angle of view.
In your first question you seem to be confusing cropping with physical size. If you want to print an 8x10 at 300dpi, you will need a file with a physical (pixel) size of 2400x3000. You can get there by cropping a larger file or by interpolating ("up-rezzing") a smaller one.
As was suggested, the actual number of dots needed for a "sharp" appearance is pretty subjective and depends on the viewing distance and your personal judgment on what's acceptable. 200-300 is a good ballpark that most viewers will consider sharp.
Thank you all for your excellent responses. I very much appreciate you all taking the time to help me sort this out. I think I have the answers now. Here’s the situation, as I understand it:
i. Clarification of what is meant by magnification: First of all, by “crop” I mean “to remove the un-needed portion of an image”. I see quite a bit of discussion in the published literature on the 35 mm format 24mmx36 mm sensor compared to the usual DSLR APS format sensor of about 15 mm x23 mm. Dan referred to a “cropped sensor” in his comments and I think that he was referring to a sensor of reduced size cropping the image. Even if that is the case, his comments were still very helpful but I need to apply them to the problem of predicting the size of a bird on my computer screen obtained by using my DSLR. The point he made (and reiterated by Guy) was that (by definition) magnification is not affected by sensor size. Magnification is defined at only one distance within the lens and camera: the minimum focusing distance. This is related to a specific distance to the subject. So, that coincides with Guy’s comment of the importance of focal length and subject distance. That means, if I am photographing a 60 mm bird with my 0.2 magnification lens, then the bird will coincide to 0.2 x 60=12 mm on the sensor regardless of whether the sensor is a reduced size sensor or a full size (35 mm equivalent) sensor.
ii. Using the magnification to anticipate the size of the cropped image on my computer screen: Now, let me try to use this information. The actual size of my camera’s sensor is 22.2mm x 14.8 mm or 329 sq. mm and the corresponding raw image from this sensor on my computer screen is 4272 x 2848 pixels or 12.2 mega pixels in total. So, I can say that 22.2 mm on my sensor corresponds to 4272 pixels on my screen (or (4272/22.2)=192.4 pixels on my screen per mm of sensor) [If I say that 14.8 mm on my sensor corresponds to 2848 pixels I again obtain 192.4 pixels per mm] and the area of 329 sq. mm corresponds to 12.2 megapixels (or (12.2 megapixels/329 sq. mm) = 37,082 pixels per sq. mm) So, now I can go from the actual size of the bird (in the field, in mm, for the distance corresponding to maximum magnification) to the size of the bird on the sensor in mm (0.20 x actual size of the bird in the field) to the size of the bird (in pixels) on my screen (=192.4 x size of the bird on the sensor=192.4x.20xactual size of the bird in mm = 38.48 x actual size of the bird in the field in mm). So, if the bird is 60 mm in height, say I want to crop a rectangle around it 80 mm high and 50 mm wide (like enclosing the bird in a frame using the actual dimensions of the bird) then I would have a crop rectangle on my computer screen which is 38.48x80=3078 pixels high by 38.48 x 50=1,924 pixels wide This corresponds to 3,078 x 1,924=5.92 megapixels. This is the answer to the second point in my original e-mail I can now compare lenses of different magnification with respect to how large would be the final cropped image on my computer screen given the actual bird dimensions.
iii. Cropped image size and printed image size: Guy mentioned that I seemed to be confusing cropping with physical size. I think that impression originated from me not distinguishing pixels from dots. Although the printed image will consist of dots, my understanding is that pixels per inch is still preferred as a method of describing resolution on the printed page because of the uncertainty as to what is meant by a dot. The 200-300 numbers that we both quote is pixels per inch, not necessarily dots per inch. [See http://en.wikipedia.org/wiki/Pixels_per_inch and the references I quoted in my initial posting].
So, if I take the higher number, 300 pixels per inch of printed photograph and I want a 4 inch x 6 inch print then I must have a cropped rectangle on my screen of 300x4=1200 pixels wide by 300x6=1,800 pixels high (i.e. I need 1200 x 1800 =2.16 megapixels for my cropped rectangle enclosing the bird). So, this answers the first point in my previous post. I can now estimate the adequacy of the cropped rectangle on my screen for a given size printed photo. However, as many of you pointed out, this is not the whole story.
The idea of “up-rezzing” , “upsizing” , “enlarging/resampling” mentioned by Guy, Gene and Howard all point to something I did not consider: using software to improve the situation. Also, it is strange but true that as Howard and Guy point out, if you are printing a larger image you actually need fewer pixels per inch than if you are printing a smaller one because of the distance to the observer. Both Gene and Howard emphasized sharpness of the image. I see that echoed in previous posts on the topic of cropping here too. Lens quality no doubt is important then. However, as both Gene and Howard point out, there are plenty of other contributing factors to sharpness too: notably camera motion and subject motion. A large, heavy lens may have wonderful optics giving sharp images but holding the lens steady during the shot could lessen the advantage (especially if I’m hiking with a monopod, for example).
From my viewpoint, all this information makes me think that consideration of cropping is something we should be paying plenty of attention to when we select a lens. It also influences our camera selection: more megapixels are better than less. Also, since cropping is also a very important part of composing the image, we may be unduly restricting our composition options if we don’t consider what a realistic crop should be for our particular situation and instead just try to minimize cropping.
The increase in megapixel capacity of cameras along with the development of improved uprezzing software makes this whole situation quite dynamic.
I am a newcomer to all of this. I hope that I’ve interpreted all of your comments correctly. If I haven’t, if you could kindly correct me once again, I will try again. Thank you.
Steve, when in doubt, get the shot - worry about composition and enhancements later. Concentrate on shooting technique. Get the best lens you can justify, 10-14 mp is fine on cropped sensor, a little more on full sensor. Make sure you carefully expose your images so you don't blow highlights or clip colors. Sharpen your images carefully, and respectful of image content. Keep in mind that a a subject shot in contrasty light will appear "crisper" and "sharper" than the same image shot in shadowless light. If you captured the resolution in the image, your prints will be sharp. If you adhere to the advice on technique and postprocessing, there should be no reason why you wouldn't be able to get images that you can upsize, as Howard pointed out, to many times the size of the original image.
While its nice to have a thorough technical understanding of all of this, there is an overwhelming number of shooters that are virtually clueless in this respect, yet consistently produce exceptional images, and vice versa.
Gene, thank you for your very valuable comments. I’ll use them. I know practical strategy born of experience when I see it!
It makes perfect sense that it is possible to obtain fine images without all this quantitative analysis. Two things caused me to look into it:
(1) the form that I filled in for submitting my images to this website asked whether the crop was more or less than 75% of the original image. I asked myself where did the 75% number come from and what are the consequences of having a cropped image which is less than 75% of the original;
(2) my only real nature photography lens right now is the Canon EF100-400 mm, F4.5-5.6 IS USM. I like it a lot but it is a bit heavy and I’m still trying to figure out how best to get a decent image in low light and how to freeze birds in flight even in reasonable light. So, I’ve been considering buying a faster, prime lens. The Canon EF 200mm f/2.8 L II USM looks like the best candidate. It’s not too expensive, considerably faster than I have (my existing lens would be f/5.0 at 200 mm) and the reviews say it provides very high quality. My main concern is that it’s only 200mm. So, the question is whether this lens is worth purchasing for nature photography. The idea that I could crop more and not sacrifice image quality makes the 200 mm look less restrictive than at first glance.
So, I do understand that we can all do without this analysis and still potentially generate fine images. However, maybe this kind of analysis can help us generate fine images more often and less expensively.
Yeah, I wouldn't worry too much abuot the cropping percentage. Going from an 8x12 proportion to an 8x10 proportion is getting pretty close to 75%. If you are submitting to the Portfolios, that is more in case if an editor wants to buy the picture. They'll want to know about how much was cropped to see if it suitable for their purposes, A spot size image can be heavily cropped and look fine in a book. The same heavily cropped image might look horrible printed as calendar size or larger. For the image galleries, other than "heavily cropped", I don't hink anyone really cares all that much, other than to know it was heavily cropped.
With regards to buying the 200 f2.8, it is a great fast lens and extremely well suited for what is was designed for. But it is also horribly short for bird photography or even mammals.. If that is what you are seriously interested in, you probably should be looking at a 500 or 600mm lens. Even a 300 2.8 would be a better choice for that type of photography. I know, this isn't what you want to hear.
My best guess is that the lack of sharpness which you are seeing is a combination of lens speed (which you already know), missed focus (the lens isn't the fastest focussing lens out there, but it is pretty good), and panning technique. Birds in flight are tough, even when you are good at it.
Posted by Steve Balke on June 24, 2009 at 10:36 PM
Dan, thank for your comments. I can see that they intuitively make plenty of sense. However, the whole reason I opened this topic was to figure out how to make some quantitative calculations of cropping and then go on to use that information. I think that these calculations reveal that the situation in lens selection etc. is changing because of all the hardware and software developments associated with the power of cropping to give satisfactory final prints from surprisingly small image areas. It may be counter-intuitive based on earlier technology. My impression is that your comments would be the same previous to this whole discussion as they are now! It may be that you don’t consider such calculations useful. Maybe they need to be expanded in some way. Maybe they are not useful for some reason. What do you think? Is it worthwhile to attempt to quantify cropping? [You’re right: I was hoping you would like the 200 f/2.8 lens more. I may need to turn to more cat photography….]
Posted by Dan Borzynski on June 25, 2009 at 12:25 AM
You are right, my comments probably would be very similar.
I have nothing against trying to come up with calculations about cropping and what would work the best. I'm all for formulas and technical specs, having spent way too many years being trained as a research mathematician. The golden ratio, a good guide for composition, I can derive in enough ways to make even me run off screaming. I'm a geek and I know it!
Having said that, it is my feeling that you are trying to overthink and overquantify the situation. The results of cropping are going to be very subject, image and camera dependent. Yes, the camera plays a role here since I have more confidence in the image quality from my 5D than my 20D.
At some point in time, what is going to be needed is to sit down with a highly detailed prop and have it photographed from various distances. Then crop, print and see if the results are satisfactory to you and the print sizes you need. There really is no other measurement here.
When needed, I can blow up an image to about 2.5x per dimension in Photoshop and produce am image I am not too uncomfortable with selling to a client as a file or a print, if it came from my 5D. My 20D, 2x per dimension. That assumes best practice on my part and this is based on my personal standards and personal testing. At this degree of enlargement, I feel that I am reaching the limits of my technique and gear.
But even with those limits, I have two show images from my 20D that I would be comfortable printing to 40x50 inches, or even bigger. That is solely because they contain no detail. Ask me to print a highly detailed image from the 20D to that size and I will laugh. (Actually, I did once to a customer. ).
The other thing is what is going on in my head when I am actually out in the field. I'll spend the time to work myself closer to the subject, shooting along the way. What I don't want to be doing is thinking to myself "just let the subject get as big as autofocus point - then I can crop and get a good picture". In all honesty, and I think this is really the main thing I want to get across, I don't believe that this line of thought leads to the best image or to develop photographically. What I do want to be thinking is how can I get it right in camera, even if the subject is small. This takes more work and has forced me to learn to stalk and shoot. I've also missed shots as well.
I don't have anything against cropping as such. The standard 3:2 ratio isn't always the best ratio for a subject. I'll even shoot knowing that I will crop later. But it is because of the subject and the camera format constraints, not because I can't get close enough.
Here is really my best suggestions:
Get a small stuff toy with a lot of detail. Shoot it at various distances, noting how big it is relative to an autofocus point. Enlarge, crop and print and see what is acceptable for you.
With the 100-400mm lens and a 1.6 sensor body, you are at 640 mm. That is a lot of focal length and coupled with something like a 50D, does give you some cropping space. Learn how to keep that combination steady for the best possible starting point for cropping and practice panning technique.
That 200 is a great lens, but I really don't think that it is the solution you are looking for.
Posted by Steve Balke on June 25, 2009 at 12:34 AM
Thank you, Dan. You gave me plenty to think about.
Steve, I just read your response - sorry for the late post - but here is another two of my cents. If you are currently using a 100-400, which is a fine piece of glass, at 400mm, you are already at the short limit of what I would consider using for bird photography. As far as birds in flight, the tripod and head and panning technique are as important as the lens -
Take a look at http://www.flickr.com/photos/9961692@N04/
I have used a Sigma 50-500 for all of the birds in flight - at F6.3 it is not a particularly fast focusing lens, but the focusing system on my D200 is acceptably accurate enough. I use a gimbal head (Manfrotto 393) and a Feisol 3472 tripod. It is not unusual for me to use ISO 800 or higher to stop motion blur. Many of the eagle shots are slightly cropped. Image softness at longer focal lengths can be usually tracked down to panning technique and inadequate camera support.
In a perfect world with deep pockets and a sherpa to carry it all,, the ideal lenses would be the 400 F2.8 (fast but over 14 lbs, over $10k), a 500 or 600 F4 - (10 and 14 lbs respectively, also expensive), or the Sigma 300-800 F5.6, which also comes in around $10k and is a real heavyweight at 35 lbs. The 400 2.8 with a 1.4x will give you 560 F4, and still be extremely sharp, while giving you the option of shooting at 2.8 without the extender if you need to.
Gene, very interesting comments. I enjoyed perusing your website and I paid particular attention to your "birds in flight" shots. I can see that you did get fine looking flight images using the lens settings you mentioned. It’s amazing though how many variables enter into the results. The size of the bird, speed of the wing (vertically and horizontally), direction of flight, etc., etc. You mention certain tripods and the desirability of heavier lenses. It makes sense that these variables contribute too.
The solution depends strongly on how we define the problem. In my case I want to be able to walk long distances without stopping for very long periods. My hiking stick transforms into a unipod. But, much of the time my photos would be hand-held shots. The 400 mm telephoto is your short limit for birds. However, it’s my high limit for lens weight! My main objective is to record what I see as faithfully as I can so that I can identify and learn about everything from bugs to birds to flowers. I’m interested in seeing nature frozen while in motion (although I don’t pass up good stationary subjects either!). The still camera captures detail that I normally would not notice. Image quality is important to me because it shows this detail and it’s fun to use images to communicate. You can see that this kind of objective is different than if I wanted to obtain the best possible image regardless of portability. The open question is how well can I do. How close can my images approach those who have your viewpoint. I’m seeking methods of avoiding the bigger lens/better tripod approach: from cropping to software to whatever. My hunch is that technological developments will eventually provide better ways. Anyway, it’s fun to try. Thank you again for your help.
Steve, I went through the same decision process and came up with the Sigma 50-500, a Bogen/Manfrotto gimbal, and a Feisol 3472 for my walkaround kit. You cannot really get results as good as the Sigma with a lighter lens arrangement (either another lens, or lens/extender combination). And as far as a tripod/monopod, I quickly stopped using my monopod at anything over 400mm, due to motion blur. You can forget about handholding in most cases - you need a lot of light, peerless technique and strong muscles. After shooting a lacrosse game with an 80-200 2.8 with a 1.4x extender handheld, my arms and shoulders were shot.
One important consideration for me was that the 50-500 offers unprecedented flexibility as far as focal length is concerned which translates to fewer lenses to carry and a lighter load. I don't, but you can add extension tubes like those from Kenko, to allow many lenses to focus closer, although the 50-500 goes to 1:4 at about 300 mm just as it comes out of the box.
Gene, thank you again very much for your input. Looking at the specs for your equipment, I can see good reasons for your selection. However, I sure hope that I can converge on a less weighty solution. Your lens and tripod alone give a combined weight of 8 lbs. My Canon EF100-400 mm, F4.5 to 5.6L has image stabilization (which lets me avoid a tripod....sometimes) and weighs 3 lb.
So, I'll keep working at this. To be continued...
Steve, the VR is useful only for static subjects, as you probably have already figured out. The 50-500 is only 4 lbs and has more "reach" than the 100-400 and only slightly bigger. It's the only reason I suggested it, since there is no lighter combination for the focal length range, and I don't have to pack a second lens to cover down to 50mm. It is often the only lens I take, and I rarely miss a shot within its focal range, and I change lenses less frequentlyin the field, minimizing the opportunity for dust to land on the sensor.
A tripod is a given for consistent results at long focal lengths. The Feisol 3472 at 3.92 lbs is absolutely the lightest tripod that offers the stability required for anything over 400mm. You could get the Gitzo 3530LS at 4.1 lbs, but I found it not to be as stiff and stable. The 5541LS is stiffer, but it weighs 6.3 lbs and costs over $1200. The Feisol then falls in between the two in terms of quality, and costs only $520 - hence the decision to keep.
In practical use you will need at least a monopod to use a 400mm or longer lens. If you shoot wildlife, which flits about all the time, you will be frustrated in short order by a less than stable camera platform. Been there, done that. BTW, if I were looking for the best possible quality, I would not be using what I own. The gear decisions I made were dictated by three simultaneous college tuitions :)
Thank you, Gene, for clarifying your answers, for your recommendations and for the additional tripod and lens information. This has been a very fruitful discussion for me.
Discussion
17 repliesi. A range of 200 to 300 pixels per inch looks like a good range for prints. So, if I want to print a 4x6 inch print at 300 pixels per inch I should try for a cropped area containing 4x300x6x300=2.16 megapixels. (That looks good: my trusty Canon provides 12.2 megapixels). Of course, 4x6 isn’t all that large a print either! Is there a popular print size in nature photography? Are 300 pixels per inch enough for nature photographs based upon your experience?
ii. The question of whether or not the subject will fit into the minimum cropped area is more difficult to answer. Lens magnification seems a bit confusing to me. Dividing the focal length by 50 should give the magnification (8x for the maximum magnification of my 400 mm telephoto). However, the specification for the lens states a maximum magnification factor of 0.2X at 400mm. I understand that this is the ratio of the height (or width) of the subject on the camera sensor to the actual height (or width) of the subject . I know the size of the sensor from the camera specs. So, I could then assume a size for the subject and figure out the area of the sensor that will be occupied by the subject. However, that number will be independent of distance to the subject (which does not make sense!). I’m missing something! Any suggestions?
Here are the main references that I found (Know any more good references?]
http://flyingsamphoto.com/digital-photo-printing-megapixels-200.shtml
http://flyingsamphoto.com/digital-photo-guide-art-how-many-pixels.shtml
http://flyingsamphoto.com/digital-photo-guide-article-006.shtml
http://www.the-digital-picture.com/Canon-Lenses/Canon-Lens-Magnification-Value.aspx
Some general guidelines - start by getting the sharpest lens you can afford. More expensive telephotos are generally sharper than their less expensive brethren, and they are also faster, so you can shoot at lower ISO and/or faster shutter speeds. Make sure your camera support system is not generating vibrations which will cause motion blur, especially when the mirror slaps up and down, and establish a workflow in your editing program that preserves as much of the original image quality as possible.
For the record, I have seen some exceptional images printed to 24x36 that were shot with a Nikon D70 (6.1mp) and 16x20 from a D2H (4.1mp). There is a software product called DXO Objects Pro that in addition to doing some "magical" adjustments with respect to lens optical corrections, it also manipulates pixels to maximize image quality, especially for large print sizes.
(Generally around 3000x2000)
The image needs to have very good original sharpness, and ISO quality for the best results.
Of course, you never sharpen the image you're enlarging/resampling.
I've resampled images using GF up to 4x6 feet in size using this method, with excellent results.
Hope this helps a bit,
Howard
In your example, you use a 400 mm lens. Pick a subject that will be 1mm in the frame with a 50 mm lens. Keeping the camera in the same spot, now switch the lens to a 400 mm lens. That same subjuct will now be 8 mm in the frame - the 8x.
Now, let's get back to the lens specifications. Take that same 400 mm lens and suppose that it says maximum magnification of 0.2. What this means is that if you focus that lens at its minimum focusing distance, a object that is 5 inches long in real life will appear to be 5x0.2 = 1 inch long back to the recording medium.
A cropped sensor does not matter for the second case of lens specifications. Here's why: there is a cone of light that is in perfect focus on the plane of the sensor (or film). If you could see that light, the 5 inch long object would still appear to be 1" long, or 0.2 magnification. Putting a full frame sensor back there would record all of that one inch while a cropped sensor would record maybe 0.6 " of it. Either way, it is still 0.2 magnification, just more or less of it gets recorded.
Here is where I begin to go beyond what you originally asked in (ii):
Some confusion comes in in that last paragraph with cropped sensor. You'll see that the focal length is 1.6 times as long. So shouldn't that mean that the maximum magnification is 1.6 x 0.2 = 0.32? No, that 0.2 means what is recorded on the sensor.
If you print to 8 times the sensor size, for a full frame sensor or a cropped sensor, you wind up with the same amount of magnification. However, if you print to the same print size, say an 8x10, from both a cropped sensor and a full frame sensor, the cropped sensor would show more magnification. (Remember: to get to an 8x10 from a cropped sensor, you need to enlarge the area even more than with a full framed sensor to get to the same size, since you are starting with a smaller recording medium). In this case, you would have effectively a 0.32 magnification, not a 0.2 magnification.
Think of the 0.2 as lens/recording magnification, the 0.32 as a *effective* printing magnification.
I am still trying to figure out how to respond to (i). As you said, there are a lot of variables to deal with, some of which are at odds with each other. My best suggestion is to make some prints at various croppings and sizes and see what works for you. What I find acceptable may be garbage to you.
In your first question you seem to be confusing cropping with physical size. If you want to print an 8x10 at 300dpi, you will need a file with a physical (pixel) size of 2400x3000. You can get there by cropping a larger file or by interpolating ("up-rezzing") a smaller one.
As was suggested, the actual number of dots needed for a "sharp" appearance is pretty subjective and depends on the viewing distance and your personal judgment on what's acceptable. 200-300 is a good ballpark that most viewers will consider sharp.
Guy
i. Clarification of what is meant by magnification: First of all, by “crop” I mean “to remove the un-needed portion of an image”. I see quite a bit of discussion in the published literature on the 35 mm format 24mmx36 mm sensor compared to the usual DSLR APS format sensor of about 15 mm x23 mm. Dan referred to a “cropped sensor” in his comments and I think that he was referring to a sensor of reduced size cropping the image. Even if that is the case, his comments were still very helpful but I need to apply them to the problem of predicting the size of a bird on my computer screen obtained by using my DSLR. The point he made (and reiterated by Guy) was that (by definition) magnification is not affected by sensor size. Magnification is defined at only one distance within the lens and camera: the minimum focusing distance. This is related to a specific distance to the subject. So, that coincides with Guy’s comment of the importance of focal length and subject distance. That means, if I am photographing a 60 mm bird with my 0.2 magnification lens, then the bird will coincide to 0.2 x 60=12 mm on the sensor regardless of whether the sensor is a reduced size sensor or a full size (35 mm equivalent) sensor.
ii. Using the magnification to anticipate the size of the cropped image on my computer screen: Now, let me try to use this information. The actual size of my camera’s sensor is 22.2mm x 14.8 mm or 329 sq. mm and the corresponding raw image from this sensor on my computer screen is 4272 x 2848 pixels or 12.2 mega pixels in total. So, I can say that 22.2 mm on my sensor corresponds to 4272 pixels on my screen (or (4272/22.2)=192.4 pixels on my screen per mm of sensor) [If I say that 14.8 mm on my sensor corresponds to 2848 pixels I again obtain 192.4 pixels per mm] and the area of 329 sq. mm corresponds to 12.2 megapixels (or (12.2 megapixels/329 sq. mm) = 37,082 pixels per sq. mm) So, now I can go from the actual size of the bird (in the field, in mm, for the distance corresponding to maximum magnification) to the size of the bird on the sensor in mm (0.20 x actual size of the bird in the field) to the size of the bird (in pixels) on my screen (=192.4 x size of the bird on the sensor=192.4x.20xactual size of the bird in mm = 38.48 x actual size of the bird in the field in mm). So, if the bird is 60 mm in height, say I want to crop a rectangle around it 80 mm high and 50 mm wide (like enclosing the bird in a frame using the actual dimensions of the bird) then I would have a crop rectangle on my computer screen which is 38.48x80=3078 pixels high by 38.48 x 50=1,924 pixels wide This corresponds to 3,078 x 1,924=5.92 megapixels. This is the answer to the second point in my original e-mail I can now compare lenses of different magnification with respect to how large would be the final cropped image on my computer screen given the actual bird dimensions.
iii. Cropped image size and printed image size: Guy mentioned that I seemed to be confusing cropping with physical size. I think that impression originated from me not distinguishing pixels from dots. Although the printed image will consist of dots, my understanding is that pixels per inch is still preferred as a method of describing resolution on the printed page because of the uncertainty as to what is meant by a dot. The 200-300 numbers that we both quote is pixels per inch, not necessarily dots per inch. [See http://en.wikipedia.org/wiki/Pixels_per_inch and the references I quoted in my initial posting].
So, if I take the higher number, 300 pixels per inch of printed photograph and I want a 4 inch x 6 inch print then I must have a cropped rectangle on my screen of 300x4=1200 pixels wide by 300x6=1,800 pixels high (i.e. I need 1200 x 1800 =2.16 megapixels for my cropped rectangle enclosing the bird). So, this answers the first point in my previous post. I can now estimate the adequacy of the cropped rectangle on my screen for a given size printed photo. However, as many of you pointed out, this is not the whole story.
The idea of “up-rezzing” , “upsizing” , “enlarging/resampling” mentioned by Guy, Gene and Howard all point to something I did not consider: using software to improve the situation. Also, it is strange but true that as Howard and Guy point out, if you are printing a larger image you actually need fewer pixels per inch than if you are printing a smaller one because of the distance to the observer. Both Gene and Howard emphasized sharpness of the image. I see that echoed in previous posts on the topic of cropping here too. Lens quality no doubt is important then. However, as both Gene and Howard point out, there are plenty of other contributing factors to sharpness too: notably camera motion and subject motion. A large, heavy lens may have wonderful optics giving sharp images but holding the lens steady during the shot could lessen the advantage (especially if I’m hiking with a monopod, for example).
From my viewpoint, all this information makes me think that consideration of cropping is something we should be paying plenty of attention to when we select a lens. It also influences our camera selection: more megapixels are better than less. Also, since cropping is also a very important part of composing the image, we may be unduly restricting our composition options if we don’t consider what a realistic crop should be for our particular situation and instead just try to minimize cropping.
The increase in megapixel capacity of cameras along with the development of improved uprezzing software makes this whole situation quite dynamic.
I am a newcomer to all of this. I hope that I’ve interpreted all of your comments correctly. If I haven’t, if you could kindly correct me once again, I will try again. Thank you.
While its nice to have a thorough technical understanding of all of this, there is an overwhelming number of shooters that are virtually clueless in this respect, yet consistently produce exceptional images, and vice versa.
It makes perfect sense that it is possible to obtain fine images without all this quantitative analysis. Two things caused me to look into it:
(1) the form that I filled in for submitting my images to this website asked whether the crop was more or less than 75% of the original image. I asked myself where did the 75% number come from and what are the consequences of having a cropped image which is less than 75% of the original;
(2) my only real nature photography lens right now is the Canon EF100-400 mm, F4.5-5.6 IS USM. I like it a lot but it is a bit heavy and I’m still trying to figure out how best to get a decent image in low light and how to freeze birds in flight even in reasonable light. So, I’ve been considering buying a faster, prime lens. The Canon EF 200mm f/2.8 L II USM looks like the best candidate. It’s not too expensive, considerably faster than I have (my existing lens would be f/5.0 at 200 mm) and the reviews say it provides very high quality. My main concern is that it’s only 200mm. So, the question is whether this lens is worth purchasing for nature photography. The idea that I could crop more and not sacrifice image quality makes the 200 mm look less restrictive than at first glance.
So, I do understand that we can all do without this analysis and still potentially generate fine images. However, maybe this kind of analysis can help us generate fine images more often and less expensively.
With regards to buying the 200 f2.8, it is a great fast lens and extremely well suited for what is was designed for. But it is also horribly short for bird photography or even mammals.. If that is what you are seriously interested in, you probably should be looking at a 500 or 600mm lens. Even a 300 2.8 would be a better choice for that type of photography. I know, this isn't what you want to hear.
My best guess is that the lack of sharpness which you are seeing is a combination of lens speed (which you already know), missed focus (the lens isn't the fastest focussing lens out there, but it is pretty good), and panning technique. Birds in flight are tough, even when you are good at it.
I have nothing against trying to come up with calculations about cropping and what would work the best. I'm all for formulas and technical specs, having spent way too many years being trained as a research mathematician. The golden ratio, a good guide for composition, I can derive in enough ways to make even me run off screaming. I'm a geek and I know it!
Having said that, it is my feeling that you are trying to overthink and overquantify the situation. The results of cropping are going to be very subject, image and camera dependent. Yes, the camera plays a role here since I have more confidence in the image quality from my 5D than my 20D.
At some point in time, what is going to be needed is to sit down with a highly detailed prop and have it photographed from various distances. Then crop, print and see if the results are satisfactory to you and the print sizes you need. There really is no other measurement here.
When needed, I can blow up an image to about 2.5x per dimension in Photoshop and produce am image I am not too uncomfortable with selling to a client as a file or a print, if it came from my 5D. My 20D, 2x per dimension. That assumes best practice on my part and this is based on my personal standards and personal testing. At this degree of enlargement, I feel that I am reaching the limits of my technique and gear.
But even with those limits, I have two show images from my 20D that I would be comfortable printing to 40x50 inches, or even bigger. That is solely because they contain no detail. Ask me to print a highly detailed image from the 20D to that size and I will laugh. (Actually, I did once to a customer.
The other thing is what is going on in my head when I am actually out in the field. I'll spend the time to work myself closer to the subject, shooting along the way. What I don't want to be doing is thinking to myself "just let the subject get as big as autofocus point - then I can crop and get a good picture". In all honesty, and I think this is really the main thing I want to get across, I don't believe that this line of thought leads to the best image or to develop photographically. What I do want to be thinking is how can I get it right in camera, even if the subject is small. This takes more work and has forced me to learn to stalk and shoot. I've also missed shots as well.
I don't have anything against cropping as such. The standard 3:2 ratio isn't always the best ratio for a subject. I'll even shoot knowing that I will crop later. But it is because of the subject and the camera format constraints, not because I can't get close enough.
Here is really my best suggestions:
Get a small stuff toy with a lot of detail. Shoot it at various distances, noting how big it is relative to an autofocus point. Enlarge, crop and print and see what is acceptable for you.
With the 100-400mm lens and a 1.6 sensor body, you are at 640 mm. That is a lot of focal length and coupled with something like a 50D, does give you some cropping space. Learn how to keep that combination steady for the best possible starting point for cropping and practice panning technique.
That 200 is a great lens, but I really don't think that it is the solution you are looking for.
Take a look at http://www.flickr.com/photos/9961692@N04/
I have used a Sigma 50-500 for all of the birds in flight - at F6.3 it is not a particularly fast focusing lens, but the focusing system on my D200 is acceptably accurate enough. I use a gimbal head (Manfrotto 393) and a Feisol 3472 tripod. It is not unusual for me to use ISO 800 or higher to stop motion blur. Many of the eagle shots are slightly cropped. Image softness at longer focal lengths can be usually tracked down to panning technique and inadequate camera support.
In a perfect world with deep pockets and a sherpa to carry it all,, the ideal lenses would be the 400 F2.8 (fast but over 14 lbs, over $10k), a 500 or 600 F4 - (10 and 14 lbs respectively, also expensive), or the Sigma 300-800 F5.6, which also comes in around $10k and is a real heavyweight at 35 lbs. The 400 2.8 with a 1.4x will give you 560 F4, and still be extremely sharp, while giving you the option of shooting at 2.8 without the extender if you need to.
What are you using for camera support.
The solution depends strongly on how we define the problem. In my case I want to be able to walk long distances without stopping for very long periods. My hiking stick transforms into a unipod. But, much of the time my photos would be hand-held shots. The 400 mm telephoto is your short limit for birds. However, it’s my high limit for lens weight! My main objective is to record what I see as faithfully as I can so that I can identify and learn about everything from bugs to birds to flowers. I’m interested in seeing nature frozen while in motion (although I don’t pass up good stationary subjects either!). The still camera captures detail that I normally would not notice. Image quality is important to me because it shows this detail and it’s fun to use images to communicate. You can see that this kind of objective is different than if I wanted to obtain the best possible image regardless of portability. The open question is how well can I do. How close can my images approach those who have your viewpoint. I’m seeking methods of avoiding the bigger lens/better tripod approach: from cropping to software to whatever. My hunch is that technological developments will eventually provide better ways. Anyway, it’s fun to try. Thank you again for your help.
One important consideration for me was that the 50-500 offers unprecedented flexibility as far as focal length is concerned which translates to fewer lenses to carry and a lighter load. I don't, but you can add extension tubes like those from Kenko, to allow many lenses to focus closer, although the 50-500 goes to 1:4 at about 300 mm just as it comes out of the box.
So, I'll keep working at this. To be continued...
A tripod is a given for consistent results at long focal lengths. The Feisol 3472 at 3.92 lbs is absolutely the lightest tripod that offers the stability required for anything over 400mm. You could get the Gitzo 3530LS at 4.1 lbs, but I found it not to be as stiff and stable. The 5541LS is stiffer, but it weighs 6.3 lbs and costs over $1200. The Feisol then falls in between the two in terms of quality, and costs only $520 - hence the decision to keep.
In practical use you will need at least a monopod to use a 400mm or longer lens. If you shoot wildlife, which flits about all the time, you will be frustrated in short order by a less than stable camera platform. Been there, done that. BTW, if I were looking for the best possible quality, I would not be using what I own. The gear decisions I made were dictated by three simultaneous college tuitions :)