10D Sharpness

[RichardtheSane]

Hi,

Firstly, great forum here - lots of knowledge :)

I have recently purchased an EOD 10D to go with my 35mm gear. My first trip out with the 100-400 IS was great, fantastic, sharp images. Second trip, some soft looking images with the same lens and similar shooting condtions. I am going to experiment further but I was wondering if anyone else had come across a similar problem or is all the paranoia I have read about soft 10D images getting to me now?

Thanks in advance
:)

[jmublueduck]

most DSLR images come out a bit soft. In fact, I've turned all in-camera sharpening off on my D60 & do everything in Photoshop.

go download a sharpening action such as the ones @ www.fredmiranda.com/actions/ -- run your 'sharp' set of images through one of the actions & you'll be surprised at the results.

nothing to be worried about.

[PaulB]

Could be paranoia. But of course that doesn't mean that there isn't a problem.

Review everything which you could have changed from one set of pics to the next.
Pay particular attention to ISO, shutter speed, focal lengths - check in EXIF info.

Don't forget the rule of thumb for shutter speed:
Shutter speed at which you can handhold = focal length of lens.

In practice it only really works for shorter lenses and the 10D has that 1.6 multiplier..............400 x 1.6 handheld = camera shake.

I shoot motorsport and use a 300/2.8 as my 'standard' lens = 480/2.8 on 10D, and would not dream of NOT normally using a monopod at 1/500th. sec for head-on shots.

If you haven't got a monopod for that 100-400 now is perhaps the time to invest in a good one.

[jmublueduck]

good point paul.

[gmitchel]

jmublueduck wrote:
most DSLR images come out a bit soft. In fact, I've turned all in-camera sharpening off on my D60 & do everything in Photoshop.

go download a sharpening action such as the ones @ www.fredmiranda.com/actions/ -- run your 'sharp' set of images through one of the actions & you'll be surprised at the results.

nothing to be worried about.

Please avoid purchasing actions from FredMiranda.com.

As artists, we should be very concerned about censorship, and FredMiranda.com encourages censorship.

FredMiranda.com is a heavily censored site. Even trying to discuss the issue of censorship in the appropriate forum results in threads that are locked and deleted.

Please buy your PS actions from sites that are tolerant and encourage free expression!

Cheers,

Mitch

[Pekka]

Few points about sharpness:

1. Use as high shutter speeds as you can. At least 2-3x what "rule" says. I use 1/4000 whenever if it is possible :)

2. Remember that with long teles you get a lot of atmosphere between lens and subject. So you may need to fix contrast later when weather is foggy or dusty. Also heat raises up from ground and that will distort air, too.

3. If in doubt, compare original size photos with other people with same lens and camera. If you see you have somehow worse results and you can rule out basic user errors (good holding technique, enough shutter speed and not the most open aperture) then you can read what I say about 10D focus in http://photography-on-the.net/forum/showthread.php?t=10526#49694 . BTW: the validity of the test method used in that message is confirmed to be good by Canon's Technical Director Mr. Westfall ( see http://www.robgalbraith.com/ubbthreads/showflat.php?Cat=&Board=UBB8&Number=139009&page=0&view=collapsed&sb=5&o=&fpart=1 )

[RichardtheSane]

Thanks for your replies, looking at the images more closely some have been taken at different times of the day and consequently shutter speeds have been slower on those. I don't think is is atmospheric conditions with the long tele lens as I usually shoot from distances of between 2.5 and 10 metres.
I think I will have to try the focus test posted, as on a few of the 'less' sharp wildlife shots I have noticed that while the small furry mammel is not so sharp, the scenery an inch or so in front of it is pin sharp. My just be me getting used to the focus, may not. I hope so.

[daveh]

Are you using a tripod? If your DOF is only an inch then you could be rocking back and forth that much while hand-holding.

[Pekka]

daveh wrote:
If your DOF is only an inch then you could be rocking back and forth that much while hand-holding.


Good point. I have noticed I have tendency to lean forwards slightly when composing, after focusing. This kind of behaviour is something everyone should watch in daily basis and try to "un-learn" it before it becomes a habit.

[SteveCliff]

Pekka wrote:
I have noticed I have tendency to lean forwards slightly when composing, after focusing.

Aha! Thank God someone else does this as well ... I thought I was the only weird one ... ;)

[RichardtheSane]

That is a very good point, I don't use a tripod and I will have to check to see if I do that next time I am out. It does figure though. Nice one.

I am pleased to say there is no focusing fault with my 10D, tried a couple of the tests around the net and all these proved sucessful. Also looking at some of my images where I was laying flat on my belly with my elbows for support (impossible to lean forward) I can see they are pin sharp every time.

I must be leaning. A new ailment that affects photographers only can now oficially be called the leans!!

THanks for your help. I'll gat a few more shots in then start putting a few up to show.

:D

[fox1]

>>Please buy your PS actions from sites that are tolerant and encourage free expression!

[BobbyC]

Fred has a great site. Don't let some disgruntled former member keep you away. I've been around there for a while and never felt restricted or seen evidence of censorship at all. Fred has some great actions for very reasonable prices, check them out.

[fox1]

BobbyC wrote:
Fred has a great site. Don't let some disgruntled former member keep you away. I've been around there for a while and never felt restricted or seen evidence of censorship at all. Fred has some great actions for very reasonable prices, check them out.



I'm just wondering why Adobe's expensive PS 7 and their USM is being upstaged by third party actions,
or is it.?

[fox1]

>>Good point. I have noticed I have tendency to lean forwards slightly when composing, after focusing. This kind of behaviour is something everyone should watch in daily basis and try to "un-learn" it before it becomes a habit.

[daveh]

fox1 wrote:
Shouldn't predictive AF take care of 'leaning' either
backward or forward.?

As far as I know, predictive autofocus (if the original poster was even using it) can reasonably handle an object moving at uniform speed. I don't believe it makes any attempt to predict the (probably random and therefore unpredictable) jerking around of the camera by the user. This is something IS does, but I believe IS only works in the other two dimensions. It seems like you could add an accelerometer to the camera to sense camera shake/lean focus problems but I don't think anyone has.

[BobbyC]

fox1 wrote:
I'm just wondering why Adobe's expensive PS 7 and their USM is being upstaged by third party actions,
or is it.?


Most of the actions use USM, just in a more complex manner than just generically applying it.

[RichardtheSane]

I was using AI focus AF, which in theory should refocus when a subject starts to move. In reality (i have found on the EOS 33 at least) that mode only responds well if the subject in focus is the only part of the composition in focus. AI servo takes a moment or so to realise you have moved a small distance and by then the shutter has been tripped :)

I will look into sharpening actions (with an open mind), but I would prefer sharpness from the beginning for obvious reasons. Thanks again for your input.

[daveh]

RichardtheSane wrote:
I will look into sharpening actions (with an open mind), but I would prefer sharpness from the beginning for obvious reasons.

If you mean for convenience at the possible expense of image quality, I understand the "obvious reasons", however I prefer to add it later to keep my options open. Without knowing something about the image, and the destination (print/screen/size) you can't really make the optimum sharpening choice.

I'm talking about "sharpening" here rather than real resolution.

[RichardtheSane]

I prefer sharpness from the beginning not for convenience, I prefer it because it gives me a better file to work with when I clean up my images in PS (which I do on all but a few images). I though that was the obvious reason, sorry if I wasn't totally clear there.
:D

[daveh]

If in-camera sharpening provides a better file than sharpening in PS then something is wrong with PS. In-camera sharpening software is usually somewhat more crude than the software used in image editors. In addition in PS, you get much greater control over the degree of sharpening allowing you to adjust for the current image and for knowledge of how the image will be used. The latter is important because you don't want to use the same sharpening parameters for on-screen and printed copies of an image.

If you over-sharpen in camera, then the damage is done and it's too late to fix it when you're using PS.

(Again, this applies to sharpening which is an image processing trick - not to resolution or focus.)

[RichardtheSane]

Thanks for your comments Daveh.
I would like to point out at this point that I do not use in camera sharpening at all. I am aiming to get the sharpest possible image recorded by the CMOS sensor so I can process it later.

I have found that my sharpness problem was down to me moving forward slightly to focus - possibly because the handling of the 10D is different to my film EOS (need battery grip :) )

[daveh]

Ah, I misunderstood you when you said "I will look into sharpening actions (with an open mind), but I would prefer sharpness from the beginning for obvious reasons."

I thought you meant you preferred sharpENING from the beginning but you had switched from sharpENING to SharpNESS and I didn't keep up.

[Tom W]

daveh wrote:


I'm talking about "sharpening" here rather than real resolution.

Its time for the big question, and its sincere: I'm having some trouble differentiating between "sharpness" and "resolute" or "in focus". I've always described, and heard people describe a very well focused picture as being "sharp". Yet, we're talking about adjusting the sharpness after the picture is taken.

I would think that if we divide a picture into pixels, as has to happen at some level, then a given pixel along the edge of an object will either be the color of the object or the color of the background depending on where it falls in the image. If I photographed a red cup sitting on a white table, I would want the fine line between red and white to be as thin and well-defined as the actual line is in real life. If the pixel was one of the red cup, I'd want it red. If the pixel was of the table, I'd want it white. From that perspective, I would want the original to be a "sharp" or "focused" and as faithful to the original scene as possible, short of some special effect that I may want to introduce.

Maybe my definition is all wrong, or maybe I'm not looking at this issue from the right perspective.

[daveh]

Sharpening is a tool used to give an illusion of higher resolution (or sharpness to be circular.) It comes in a variety of algorithms though USM is the most commonly known. USM by the way means UnSharp Masking which is actually a trick from the film darkroom days so doing tricks to create "sharpness" is nothing new. It involves shooting the final print through a combination of the original negative and an defocused negative copy.

The way it's done in software is generally quite simple - any place you see pixels that are different - you simply increase the contrast in that area. It's a trick of color.

The sharpness controls in TV sets do the same thing. They distort the image but most TV buyers like the distortion - at least in the show room where they say - oh look that one's sharper - let's get it. (Then they get it home and see what it does to their cable signal and call the cable co to complain. The cable guy, if he's "good" finds a way to distract them while turning down the sharpness and thus "fixing" the cable signal ;))

The problem with taking the red cup on the white table is that the line will hardly ever line up exactly with a line of pixels. (In fact in the case of a cup, it's round so it will NEVER line up over any significant range of pixels. Instead you'll get shades of pink in these unaligned areas.) This will happen no matter how high the resolution and when you display it at 100% it on your screen it will always have a soft appearance because you're now seeing that pink area at a very low screen resolution (70-100 dpi typically.) But if you want it to SEEM sharp, you can apply some USM. USM will see the red/pink/white transitions and make the pink darker and the white whiter. You perceive that as sharper though the dpi hasn't changed.

What's confusing is that Sharpening software makes something look sharper without doing anything about the focus, resolution, or camera shake.

[Tom W]

daveh wrote:
Sharpening is a tool used to give an illusion of higher resolution (or sharpness to be circular.) It comes in a variety of algorithms though USM is the most commonly known. USM by the way means UnSharp Masking which is actually a trick from the film darkroom days so doing tricks to create "sharpness" is nothing new. It involves shooting the final print through a combination of the original negative and an defocused negative copy.

The way it's done in software is generally quite simple - any place you see pixels that are different - you simply increase the contrast in that area. It's a trick of color.

The sharpness controls in TV sets do the same thing. They distort the image but most TV buyers like the distortion - at least in the show room where they say - oh look that one's sharper - let's get it. (Then they get it home and see what it does to their cable signal and call the cable co to complain. The cable guy, if he's "good" finds a way to distract them while turning down the sharpness and thus "fixing" the cable signal ;))

I've always thought of "sharpness" as being synonymous with "in focus". This, apparently, isn't so.

The problem with taking the red cup on the white table is that the line will hardly ever line up exactly with a line of pixels. (In fact in the case of a cup, it's round so it will NEVER line up over any significant range of pixels. Instead you'll get shades of pink in these unaligned areas.) This will happen no matter how high the resolution and when you display it at 100% it on your screen it will always have a soft appearance because you're now seeing that pink area at a very low screen resolution (70-100 dpi typically.) But if you want it to SEEM sharp, you can apply some USM. USM will see the red/pink/white transitions and make the pink darker and the white whiter. You perceive that as sharper though the dpi hasn't changed.


I wasn't really thinking of the monitor screen, per se, but I was trying to illustrate what I thought was the definition of sharpness. Yes, there will be some pixels that will be on the line between white and red. Presumably, as you say, they will be some shade of pink. Now, assuming that the number of pixels is plentiful enough to make the transition from red to pink to white, then there will appear a sharp or in-focus line.

I think I see what you're saying now - adjusting sharpness by making dark pink pixels darker, and light pink pixels lighter to make a more defined contrast or line between the red cup and the white table. I can see "oversharpening" when you get a jagged line instead of a straight line as depicted in reality. I would think that given enough pixels, sharpening wouldn't be necessary as the transition would be sufficiently small to be invisible to the eye (unless we blow the picture up enough).

What's confusing is that Sharpening software makes something look sharper without doing anything about the focus, resolution, or camera shake.



Now that makes a good deal of sense - camera shake, focus, etc are there already. An out-of-focus picture can be sharpened, but it is much more difficult to make it "in focus" just by improving the contrast along the transitional line when the line is blurred (I would guess that would be a wide band of varying pink pixels in the table/cup example). I can guess that there would be a sharpening of the blur in the picture as well. Interesting. Thanks!

[daveh]

I would think that given enough pixels, sharpening wouldn't be necessary as the transition would be sufficiently small to be invisible to the eye (unless we blow the picture up enough).

Well the conventional film USM is often applied to to medium and large format negatives which have a lot of resolution. But then when you have a lot of resolution, it just makes people want to print larger. Eventually we'll be using USM to help hide diffraction limits rather than pixel limits. (In fact that's not very far away I think.)

[Roger_Cavanagh]

daveh wrote:

I would think that given enough pixels, sharpening wouldn't be necessary as the transition would be sufficiently small to be invisible to the eye (unless we blow the picture up enough).

Well the conventional film USM is often applied to to medium and large format negatives which have a lot of resolution. But then when you have a lot of resolution, it just makes people want to print larger. Eventually we'll be using USM to help hide diffraction limits rather than pixel limits. (In fact that's not very far away I think.)

Also smaller pixels mean a worse signal to noise ratio. That's why cameras like to the D30 with a "mere" 3 million pixels give good results compared to P&S cameras with more, but smaller, pixels.

Here's a good article about sharpness: http://www.luminous-landscape.com/tutorials/sharpness.shtml

Regards,

[Tom W]

daveh wrote:


Well the conventional film USM is often applied to to medium and large format negatives which have a lot of resolution. But then when you have a lot of resolution, it just makes people want to print larger. Eventually we'll be using USM to help hide diffraction limits rather than pixel limits. (In fact that's not very far away I think.)

Printing larger has its advantages - I have a pile of 8X10's and 8X12's sitting here in various stages of framing in prreparation of making a full wall of pictures of the city in which I was raised. On the other hand, I have a couple of photo albums full of snapshots that would never end blown up to full frame. Just little memoires of family gatherings and vacations and such. Perhaps we are fortunate that common printers aren't bigger. :)

[Tom W]

Roger_Cavanagh wrote:

Also smaller pixels mean a worse signal to noise ratio. That's why cameras like to the D30 with a "mere" 3 million pixels give good results compared to P&S cameras with more, but smaller, pixels.

Here's a good article about sharpness: http://www.luminous-landscape.com/tutorials/sharpness.shtml

Regards,

Roger - thanks for the link. I'm a little curious about the signal-to-noise ratio vs. smaller pixels. I'll read some more, but I have to ask if this is a function of smaller pixels or of technological limits of acheiving smaller pixels? I think that one must also consider the smaller lens used with smaller P&S cameras like my S400 - it cannot compete with a 35-sized lens system.

This is turning out to be most informative - the more I learn, the more I find that I did not know.

[daveh]

Here's the basic problem: For any given area sensor area, more pixels means smaller pixels and smaller pixels means that each pixel receives less light. Less light means more amplification of a weak signal which means more noise.

[Tom W]

daveh wrote:
Here's the basic problem: For any given area sensor area, more pixels means smaller pixels and smaller pixels means that each pixel receives less light. Less light means more amplification of a weak signal which means more noise.


That makes sense (no pun intended). So in order to get more pixels and keep the noise level down, what we need are either larger sensors (1Ds??), more sensitive pixels, or cleaner amplifiers or a combination of all three.

[daveh]

Tom W wrote:

That makes sense (no pun intended). So in order to get more pixels and keep the noise level down, what we need are either larger sensors (1Ds??), more sensitive pixels, or cleaner amplifiers or a combination of all three.

Yes although there may be fundamental, economic (1ds?) or technological limits to each of these. It's interesting how often counterbalanced fundamental properties apply to photography. For example, you increase the format size to reduce the noise (film or digital - doesn't matter really) and then you find yourself using smaller apertures - both because you need the smaller aperture to get the same DOF on the larger format and because it's hard to scale fast lenses up. So what do you do if you want to handhold? Use faster film (or more amplification). Result: more noise. Darn it!

[Tom W]

daveh wrote:
Tom W wrote:

That makes sense (no pun intended). So in order to get more pixels and keep the noise level down, what we need are either larger sensors (1Ds??), more sensitive pixels, or cleaner amplifiers or a combination of all three.

Yes although there may be fundamental, economic (1ds?) or technological limits to each of these. It's interesting how often counterbalanced fundamental properties apply to photography. For example, you increase the format size to reduce the noise (film or digital - doesn't matter really) and then you find yourself using smaller apertures - both because you need the smaller aperture to get the same DOF on the larger format and because it's hard to scale fast lenses up. So what do you do if you want to handhold? Use faster film (or more amplification). Result: more noise. Darn it!


Darn it is right! As for noise, that's a matter for the fast-advancing technology of sensors and electronics to resolve. I think larger sensors make sense for those of us with full-sized lenses (compatibility). I also think that we'll see a Canon full-sized sensor digital for under $2000 within a year and a half (but that's just a wild guess, or maybe wishful thinking).

Right now, I'd love to give the 1Ds a test drive, but I fear that economics is my limiting factor on driving one of those home! But that will change. In the meantime, I'll plod along with the Elan, the S400, and the occasional use of the old FT (which still works quite well, despite a very dusty viewfinder). Heck, I might even put a roll into my father's old Argus!! :)

[daveh]

Tom W wrote:
I also think that we'll see a Canon full-sized sensor digital for under $2000 within a year and a half (but that's just a wild guess, or maybe wishful thinking).


Perhaps but Moore's law which has been the driving force in semiconductor performance increases and price decreases since 1965 is all about squeezing more elements onto the same amount of silicon. A full-size sensor should, in theory, drop in price far more slowly than other electronics because it doesn't have the advantage of getting smaller. It has to rely only on process and productivity improvements to get cheaper.

Then again, Canon seems to want to "own" the DSLR market very badly. Maybe to the point of selling bodies at a loss so they can sell you lenses for the rest of your life.

[RichardtheSane]

Some very good reading in this thread, I have learned a lot - including how much more there is to learn :D

[Tom W]

daveh wrote:
Tom W wrote:
I also think that we'll see a Canon full-sized sensor digital for under $2000 within a year and a half (but that's just a wild guess, or maybe wishful thinking).


Perhaps but Moore's law which has been the driving force in semiconductor performance increases and price decreases since 1965 is all about squeezing more elements onto the same amount of silicon. A full-size sensor should, in theory, drop in price far more slowly than other electronics because it doesn't have the advantage of getting smaller. It has to rely only on process and productivity improvements to get cheaper.

Then again, Canon seems to want to "own" the DSLR market very badly. Maybe to the point of selling bodies at a loss so they can sell you lenses for the rest of your life.

Really, I think mass production (productivity) will bring the price of full-sized sensors (or any sized sensor if it becomes popular) down to where they can be manufactured as cheaply as a Pentium-2, and I don't think it will take all that long. Also, Moore's law notwithstanding, I think there's still some good R&D going on with some significant gains to be made. True, advances will become less and less significant, but there will be advances. And, this is new ground in some ways - the application of electronics in this manner. Watch for the next generation of the Digic processor - I expect a significant improvement there as well, even if only from improved application of existing technology.

[daveh]

A Pentium 2 is a MUCH smaller chip. (As is are the 3 and 4.) By the way, the new process is CMOS - which is to say it's the old process. (ie same as most things that aren't sensors.)