ragebot wrote:
Wow I am having trouble with copying and pasting so I am just gonna start fresh.
Dick Lyons has a nice video some where on the web about pixels that explains things very well, but I cant seem to find the link just now. It was in a talk he gave at Cal Tec.
One of his points is that not all pixels are created equal. Clearly an 8 bit file has fewer colors than a 64 bit file; but there is more to it than that. A Bayer sensor is guessing what the data about 60% of its detectors collected because of the mask. Combined with the less than perfect optics and software issues makes the roll off even worse.
But even if we grant the best optics and best exposure there is another issue in the Nikon v Canon stuff. For what ever reason there are lots of posts about how Nikon uses more aggressive in camera sharpening and lower pixel density to produce sharper straight out of the camera images while Canon uses denser pixels and less in camera sharpening to allow more detail to be brought out with effective post processing.
One thing I have noticed is that with a 1d4 I seem to use the unsharp mask a lot less than with my 1d2; but still more than with my Sigma bodies. I also seem to be cropping much more at 100% crops with the 1d4 than the 1d2.
There have also been tons of threads about the MP wars and how some of the P&S digi cams have silly MP numbers, if the IQ of the images they produce are used as a judge.
While we like to think science always has the answer I sometimes thing we forget how much art fits into the picture. There is a trade off between pixel density (the d3 has great low light ability, while the 1d4 has more detail), file size, burst rate, and many other factors in creating a DSLR.
The 1d4 allows me to consistently capture images with IQ that meets my standards; more so than any other body I have. If another body could do that I would buy it....Show more →
i think you are overstating how bad the roll-off is
also if you compare a bayer to a foveon of similar cost and total bits to process the bayer images generally tend to look a lot better overall
a lossless 640x480 video looks better than a modestly compressed one of the same dimensions (and smaller fizlesize) but a modestly compressed one of same file size and larger dimensions looks better than the much smaller lossless video
skibum5 wrote:
i think you are overstating how bad the roll-off is
also if you compare a bayer to a foveon of similar cost and total bits to process the bayer images generally tend to look a lot better overall
a lossless 640x480 video looks better than a modestly compressed one of the same dimensions (and smaller fizlesize) but a modestly compressed one of same file size and larger dimensions looks better than the much smaller lossless video
I really don't understand your post.
Roll off is not constant, it changes from image to image.
I have no idea what you mean by total bits, and the cost of cameras also changes rather quickly.
I don't recall posting anything about video in this thread.
In regards to motion blur and focal length or sensor size, I agree with Nill that in most cases motion blur is not affected by these. If you are shooting a soccer player with a 100mm lens at 100 feet with a full size sensor, and the player fills the frame from top to bottom, and the player can cover half the horizontal frame is say 2 seconds, then the blur will be the same when you switch lenses provided the player still fills the same amount of the frame.
So if you want a frame filling shot at 160mm with the player filling the frame from top to bottom, then you will need to step back to 160 feet to have the same magnification of the player, and the blur will be the same. The player will still be covering the same number of feet in that 2 seconds.
Conversely, you can keep your 100 lens and shoot at 160 feet and the player will only be 0.625 times (100 feet/160 feet) as large in the frame. Then you will need to enlarge the image by 1.6 times to get to the same print size.
Same for a 1.6 crop sensor. You will need to step back to 160 feet or you will need to decrease your zoom from 100mm to 62.5mm.
So when you a shooting with the attempt to maintain a certain image size in the viewfinder (100% view), the motion blur will be the same for any lens or sensor since the image magnification has not changed. The player is still covering half the frame in 2 seconds.
In other words, subject motion blur, as a proportion of the subject's size, is independent of lens and pixel density, and only depends on shutter speed. Isn't that obvious? Sometimes I think this site is just a wrong maths death-match.
There is an advantage in a camera with a higher pixel density in that it is capable of capturing and resolving finer detail than a camera with a lower pixel density. The trouble is that our definition of detail will vary for different subject matter. If we contrive a case where the detail we are photographing is made up of fine or narrow lines, rather than edges between two large areas of colour or tone, then there is a point at which a high-res camera will capture that fine detail at noticeably higher contrast than a camera with bigger pixels. This is because the fine details may completely fill the smaller pixels but not fill the larger pixels. Now lets consider the effect of motion blur in which the extent of the blur is large compared to the size of the details. The high-contrast image captured by the high-res camera will fall to pot because none of the pixels are capturing the detail for the entire duration of the exposure; contrast will fall off significantly and the image will look quite blurry. It would look blurry on the low-res camera too, but because it started out blurry (low in contrast) it is harder to see the deterioration than it is with the high-res camera.
This is a contrived example but it shows that motion can cause a significant loss of perceived detail when the combination of pixel size, detail size and blur size is right. The sort of details I have described could be lines in artwork, blades of grass in a landscape, bits of feather on a bird, etc.
I think that if you were looking at large details such as the team numbers on a player's shirt then the details at pixel level are more likely to be seen as single edges between two relatively large areas of colour, and this single edge detail will tolerate a greater degree of motion blur.
Nill, earlier you mentioned an example in which the lens is bumped say 1/8 inch and concluded that the camera blur is different to the subject motion blur. If you translate that bump into a sideways movement of camera and lens with everything moving the same amount then I think you are correct. However, if you translate the bump into an angular rotation rather than a sideways movement then it would be the same effect as subject motion regardless of distance to the subject.
As a matter of interest, IS also assumes that any detected movement is angular rather than lateral, and corrects accordingly. It works for any subject distance. However, if the lens was moved without changing the angle of the optical axis then IS would not make a proper correction. This is why IS was considered inappropriate for macro lenses. The new IS macro lens must do things a bit different unless it also does not work properly at close distances.
Alan, I take your point, and it's a good one, but I'm inclined to think you have it backwards. That is to say, that lateral displacement of the camera and lens is effectively the same as subject motion, regardless of distance, whereas angular displacement is different because the the effect of the angle and the distance is to magnify the effect on the subject.
In fact, perhaps we should describe three categories of motion leading to greater or lesser amounts of motion blur:
1. Subject motion;
2.a. Lateral motion of camera and lens in the same plane relative to the subject;
2.b. Angular motion of camera and lens relative to the subject;
If during the exposure we move the camera/lens perfectly sideways one inch, in the same plane relative to the subject, it's pretty much the same as if the subject (or more correctly, the entire contents of the frame) had moved that same one inch.
But if instead we bump the lens an inch such that it rotates around the axis of the camera (which is likely, working on a monopod as I do), that one inch gets magnified by the change in angle and the distance to subject... it forms a big triangle.
Imagining larger displacements makes this easier to visualize. Say I put my 400mm on a tripod and point it at a stationary target 50 yards away. Now I carefully slide the whole rig six inches to the left, without changing the angle at all. What I see in the viewfinder has changed, but only by six inches — indeed the very same as if my subject (and of course the background and everything else) had slid to the right by six inches.
But if instead of sliding the entire rig sideways, I move the lens such that the front element moves six inches to the left, rotating around the axis of the ballhead's pan base, what I see now through the viewfinder has changed by a hugely different amount. I've moved the angle perhaps 30 degrees (just grabbing a number without actually measuring and calculating, mind you), and now I'm not even looking at the same scene. I'm looking instead at an entirely different scene, many yards to the left of the original frame.
The bad news is I suspect as a practical matter most of the camera motion we have to deal with, especially shooting from a monopod rather than handheld, is the more troubling angular displacement as opposed to the somewhat more benign lateral movement. Shooting handheld it may be less so (relatively speaking, in that there's more motion of the camera body as well as the front element), but I would bet that in general terms there's still more angular than lateral motion — all of which would be consistent with IS being designed to expect the former.
Nill
p.s. As I think it through further, I believe perhaps 1. and 2.a. are essentially the same, except insofar as greater enlargement comes into play. A lateral inch is only an inch, whether it's motion of camera/lens or of subject. And distance to subject matters with respect to lateral motion of either camera or subject only if and to the extent we choose to enlarge the subject to a greater degree, thus similarly enlarging the visible effects of that one inch of motion.
p.p.s. To put this into even more practical terms... I have a terrible tendency when shooting field sports on a monopod (which I do a lot) to jig the camera sideways when I mash on the * button to AF. All this recent discussion of motion blur has made me even more acutely aware of this very significant flaw in my technique. I always rest my left hand on top of the lens, but to alleviate this personal idiosyncracy I have lately taken to moving that hand out farther on the lens, right out to the front edge of the lens hood. This helps somewhat to control the sideways motion. Even better would be to learn to hit the * button more gently, and with more of a squeeze than a press. (This is all rather like shooting a gun, isn't it? But I haven't fired a gun since I was eight years old.)
Our exchange above has put me in mind of another adjustment I might try. I'm thinking if I slide the QR plate back farther in the monopod's clamp, such that the amount of lens in front of the axis of rotation is reduced and the amount behind increased, that might serve to mitigate at least slightly the effects of my abominable camera handling skills.
