Deep depth of field to get maximum detail on human faces

[Eildosa]

Hello,

I'm quite new from photography, basicaly I started learning yesterday.

Let me give you some context.

I have to make pictures of human faces maximizing informating (sharp picture, no noise, bigger resolution is better) in order to make software 3D reconstruction.
ISO 100 is mandatory, I'm taking 21 pictures, if one camera add noise it won't match with the noise of the other ones, generating artifact in 3D reconstruction.
Obviously you cannot totally eliminate noise, but you can minimize it with the lower ISO setting.

Until now I was shooting in automatic mode, most of the time only 1 part of the face was in focus, the forehead, the nose, or the chin, the rest was blurred.
futhermore the 21 camera where selecting different settings, generating very bad 3D scab.

From what I understand reducing aperture, and thus increasing the f-stop number gives me a wider depth of field allowing me to get more of the face in focus.
However smaller aperture = slower shutter speed to get correct exposure and might cause movement bluriness.

I did some try with a TZ40, a lot of light and a silicone mask. the camera is 30 cm away from the face.
image stabilization off, Manual mode (apperture, shutter speed and ISO setting can be set by user).
Camera mounted on a tripod, using the 2 second timer to avoid shacking when taking the picture.

With f3.3, shutter speed of 1/25 and ISO 100 and focus made on the left cheeck I'm getting a regular picture.
With f8, shutter speed of 1/4 and ISO 100, focus made on the left cheeck the background is less blurry but the whole face is soft.

why is that? I read that small aperture tend to render picture soft, but this is not suposed to happen before F12.

TL;DR : why is my TZ40 generating soft pictures at F8?

Thanks.

Ps : I do not have acces to the pictures right now, I will be able to post them in 12h.
Ps : look like I posted in the wrong categorie, sorry

[NorthernFocus]

A bit more information would help. What camera/lens combination are you using. From what you wrote it sounds like you are taking the 21 shots with multiple cameras, is that correct?

Depending on the rigidity of your tripod setup, even with a 2 second delay you could still have camera shake which could be the source of the softness you describe. The shutter speed you mention of 1/4s is in a range that is particularly troublesome for effects of camera shake. If your camera has a mirror up mode, you should use that. Flip the mirror up, pause 5 seconds, then engage the shutter. Adding weight to the tripod can also help.

One comment about noise, ISO is not the only factor to consider. Forcing a low ISO setting can result in a tendency to under expose shots and under exposure is very bad for causing noise in dark areas of an image.

[truonda]

Here are the many ways you can control the depth of field:

http://media.digitalcameraworld.com/...heat_sheet.jpg

[Manfred M]

Lots of questions without a lot of information on what you are doing.

The TZ40 is a basic point & shoot, I assume with the usual tiny sensor that tends to be standard in this type of camera.

The part of the face being in focus is related to the magnification and f-stop you are shooting at PLUS the sensor size. Generally, with small sensor, one will get a deeper DoF than a larger sensor, at a given f-stop.

30cm from the face makes me suspect that you are shooting at a fairly wide angle. This will introduce both distortion and will be the reason you are seeing the DoF issues you have reported. A longer focal length, taken as some distance (several meteres / yards) away will help here.

Noise should not be a factor at 100 ISO, so long as there is enough light. I would not shoot at less than 1/60th sec to prevent motion blur of the subject (the camera on the tripod won't move, but your subject might).

[arith]

Why not try a light tent, if your getting 30 cm away the object must be small. I would say that 70mm focal length is best at a distance greater than 3 metres using a tripod and lights, then f4 gives greater than 30 cm dof on a 1.6 crop.


But this might do the job.

http://www.ebay.co.uk/itm/50-50-50-c...item415cb81aab

[Mike Buckley]

Why not try a light tent

The subject is a human face. The light tent won't work.

[Manfred M]

n 30 cm dof on a 1.6 crop.

Shots are being taken with a Panasonic TZ40, so a lot more DoF than with a crop Canon. I believe it has a tiny 1/2.3" (6.2mm x 4.6mm) sensor. Taking the shots 30cm away is way too close. 1/25th sec is way too slow (possibly part of the softness issue).

Just some educated guesses without seeing the images.

[Eildosa]

Depending on the rigidity of your tripod setup, even with a 2 second delay you could still have camera shake which could be the source of the softness you describe.

It is true that the rig is not very stable. the following pictures where done with the 10 second time.
It did not seem to move anymore at 5 second.
Do you think activating the "anti shake" functionnality would help.

One comment about noise, ISO is not the only factor to consider. Forcing a low ISO setting can result in a tendency to under expose shots and under exposure is very bad for causing noise in dark areas of an image.

Dark area are not a problem, the whole face is evenly illuminated, only the face is being processed by the software.
A mask is being applied for the outside area to be ignored.

Here are the pictures :


This picture is blurry becaused I used the HDR mode whitout a tripod

This is my rig, made of wood, 21 cameras, on the pictures you can see some nikkon coolpix L27, when I bought those I didn't know that you could not set ISO, aperture or shutter speed on low end cameras. Big mistake :/
Each camera chooses it's own iso and shutter speed, the result are unprecise 3D reconstruction.
I cabled each cameras and made a remote for them, all I can do is focus and shoot.

I put my TZ40 in the middle when doing the test, because I'd like to replace them by some canon a2500 (so I was experimenting with how different it would be if I was able to shoot in manual mode), apparently using CHDK I should be able to set the ISO, aperture and shutter speed.

Increasing distance from the subject is not possible do to available space. furthermore I would have to manually zoom with each camera, I won't be able to set exactly the same zoom value and thus will get a different focal lenght on each cameras, and if I remember correctly you cannot make a 3D reconstruction when using different focal lenght.

This is a comparaison of the 2 pictures :


Theres is a lot of light but at F8 I cannot use a shutter speed faster than 1/4 (with ISO at 100) otherwhise the picture is under-exposed, I cannot increase the light or it will hurt your eyes.
If the softness is due to micro-vibration on the rig then I am doomed to use F3.3 for live subject.

Here are the full resolution pictures :
https://dl.dropboxusercontent.com/u/...n/P1010165.JPG
https://dl.dropboxusercontent.com/u/...n/P1010173.JPG

[arith]

Sorry I thought with the closeness and long shutter speed it must be a model of a head. It's much too close and will leave perspective distortions which will look unpleasant. The slow 1/4 of a second shutter speed on a moving object is going to be blurred.

I think the only thing you can do is get a camera that can take a longer focal length lens and a couple of speedlights. Sorry I couldn't help.

[Manfred M]

Hmm. Now that I see your setup, I cannot for the life of my figure out why you would have gone through such a convuluted setup to photograph an inanimate object. An indexing turntable with a single camera on a tripod with a decent lens lit by a couple of softboxes at 45 degrees and the one right behind the mask would be a lot easier.

Your image shows a lot of distortion as you are shooting at a wide angle, the the small point and shoot cameras tend to have horrible barrel distortion that is tough to correct for, so trying to stitch that image together is simply not going to work well.

You could always shoot with a prime lens (non-zoom, fixe focal length) with an interchangeable lens camera for a far smaller investment.

[Eildosa]

No no, I only use this plastic head when there is nobody around.
The purpose is actually to use it on real human.

This kind of distortion is'nt actually a problem, the 3D reconstruction algorythm has no problem correcting it.

[Manfred M]

I can see this setup if you are shooting people.

A few more thoughts:

If the shots are soft, it is because you are much too close and the level of magnification means you do not have enough DoF to get the resolution that you need. That means moving the cameras further back, adding more light and increasing your shutter speed to at least double what it is to freeze the motion. You might have noticed, photographers use flash rather than continuous light sources for people shots.

For lighting, you are likely to find that continuous light sources are likely inadequate for what you are doing. Even small flash is running around 70W-s. A small flash fires for around 1/1000th sec. To get that kind of output at say 1/50th of a second, you need to pump in some 1400W; which is a lot more than you have in your setup. This should also kill any digital noise issues that you have

Again, if the software takes care of the distortion, I do expect some of the softness you are seeing is likely due to the quality of the optics that you are using.

[Mike Buckley]

This thread is clearly out of my league as I suspected, but I sure do find it fascinating. Time for me to return to lurk mode.

[Eildosa]

I can see this setup if you are shooting people.

A few more thoughts:

If the shots are soft, it is because you are much too close and the level of magnification means you do not have enough DoF to get the resolution that you need. That means moving the cameras further back, adding more light and increasing your shutter speed to at least double what it is to freeze the motion. You might have noticed, photographers use flash rather than continuous light sources for people shots.

For lighting, you are likely to find that continuous light sources are likely inadequate for what you are doing. Even small flash is running around 70W-s. A small flash fires for around 1/1000th sec. To get that kind of output at say 1/50th of a second, you need to pump in some 1400W; which is a lot more than you have in your setup. This should also kill any digital noise issues that you have

Again, if the software takes care of the distortion, I do expect some of the softness you are seeing is likely due to the quality of the optics that you are using.

Well I'm quite disapointed from this camera, I mean come on, 250 euros and the pictue quality is no better from the 44 euros Coolpix L27 (when they choose the right exposure setting).
With the a2500 I'l be able to set the zoom to a fixed value with a script on the sd card, putting the camera further will be possible.

I'm gonna build a flash setup with 8*100W high luminosity led, but I'm quite scared for my subject eyes.

[Manfred M]

I'm gonna build a flash setup with 8*100W high luminosity led, but I'm quite scared for my subject eyes.

I wouldn't get too concerned, back in the 1950s, all those beautiful B&W images we see of the famous Hollywood stars were taken under high-intensity "hot" lights that were real "scorchers", up in the 2000W to 3000W range. Your subjects will probably squint, but at least you won't have a heat issue.

I still think some form of flash would work better and still have some serious doubts about the camera type you are using.

[Eildosa]

I wouldn't get too concerned, back in the 1950s, all those beautiful B&W images we see of the famous Hollywood stars were taken under high-intensity "hot" lights that were real "scorchers", up in the 2000W to 3000W range. Your subjects will probably squint, but at least you won't have a heat issue.

I still think some form of flash would work better and still have some serious doubts about the camera type you are using.

The final purpose is not to get Excellent pictures, but the best I can get with this type of camera.
Current photogrammetry software are made for high quality DSLR pictures but as this paper show it is
possible to genere high quality scan from point shoot cameras.

http://www.disneyresearch.com/wp-con...nner_paper.pdf

I want to do my own reconstruction software but I also want to be able to get the best pictures from my cameras.

Someone said that softness could be because of the close range and the poor optics.
Will placing the cameras at 2 meter from the subject and using the zoom to get all the face in the pictures generate the same softness?

Well then, I'm going to build this flash.

[Downrigger]

Interesting problem, intriguing set-up and objective. For analysis of expression and emotion, maybe? I note that the cameras in the disneyresearch paper are large sensor relatively high-end DSLRs.

You might just experiment with one camera. You would find, I think, that the further from your subject and the greater your focal length, the less the distortion will be evident, while at any given distance, you can increase depth of field by decreasing focal length (then, depending on extent and requirements of your software, crop more). But then, if those maneuvers point to using a greater distance from subject, you'd have to reconstruct that fantastic lighting array and add more lumens too. In that regard, I wonder also about the tolerance of the system for noise. Have you pushed the system with higher ISOs to see what the tolerance of it is for noise. I would think you could go higher than 100, but you may know that to be the case empirically.

Thanks for sharing this interesting work.

[Manfred M]

Someone said that softness could be because of the close range and the poor optics.
Will placing the cameras at 2 meter from the subject and using the zoom to get all the face in the pictures generate the same softness

Unfortunately, it is difficult to say exactly how a specific camera works without testing it. The issue with the point & shoot cameras tends to be two-fold:

1. They have relatively long zoom ranges in a fairly compact (and inexpensive) package, when compared to higher end gear. This means that the lens designer is working against three significant constraits;

a. The overall camera dimensions;

b. The cost constraints so that the camera can be sold at the desireable price point; and

c. The wide zoom range demanded by the people that buy this style of camera.

This effectively means that there will be distortion and softness, especially at the ends of the focal range. Distortion (barrel, pin-cushion), chromatic abberations and some degree of softness tend to be the outcome.


2. The small sensor size means that any image capture defects are magnified as you increase the size of the image. A small sensor (especially with a high pixel count) means a higher succesptibility to capture noise; and the technique often used to correct for that is aptly named "pixel smearing", which will most definitely result in some noticable level of softness (assuming that you are using jpeg output from the camera).

My guess is that you should get a better shot taken somewhere towards the mid-range of the camera. Taking shots of a grid with the camera perpendicular to it, and looking at it magnified to at least 100% (probably higher) would enable you to figure out where the "sweat spot" in the camera range is found. I would suspec that a shot from 2m would be sharper than having the camera working at the widest angle.

[Eildosa]

Well I just ordered one canon a2500.
In 1 week I'll try every possible variation for ISO, shutter speed and aperture at 30cm, 1m with zoom and 2m with zoom.

Around 1500 pictures.
I'll obviously make a chdk script that operate the camera on it's own.

I'll post the interesting part of the result if you are interested.

[Manfred M]

Frankly, I rather doubt that the Canon will do anything a lot differently than the other cameras you have been working with. These low-end point & shoot cameras have pretty well equal performance optically, regardless of the brand badge on them.

I find it interesting that the Disney team (in the link you sent out) is using four high end cameras for their work, rather than a fleet of small, cheap ones. I suspect you might get identical results, with a lower cash outlay if you shot with run-of-the-mill webcams.

http://www.cameralabs.com/reviews/Ca...werShot_A2500/ no surprises in the review...