SPIE (the International Society for Optical Engineering) See:
http://spie.org/x16218.xml is holding a conference on 3D imaging from Jan. 23 – 27 in San Francisco, CA. My paper and presentation: “Human perception considerations for 3D content creation” is about the problem of perception conflicts as they relate to 3D imagery and what to do about them.
I first started thinking about this when I saw an old lenticular photograph of Queen Elizabeth. The photograph could be viewed with stereopsis but the Queen looked like she was dead. Watching the movie Beowulf, while not in 3D, also gave me the creeps as the characters had a dead aspect to them. I noticed some 3D lenticular photographs of people presented with a doll-like character. I then started to notice things in 3D movies that didn’t seem right. When details disappeared into blackness or got blown out to white I noticed an uneasy feeling while looking at that part of the 3D presentation.
Indeed, every time something was presented in 3D that was atypical or not possible to see in the real world, I could detect a feeling of conflict present at some level in my subconsious and I started to manifest a sensitivity to it with regards to recognizing when it was happening.
All of these observations got me thinking about the various mechanisims that we use to see and interpret depth, space and texture. Certainly vergence is the primary mechanism, but as I became more aware of supporting clues like accommodation, motion, luminance dynamic range, binocular rivalry, field of view and so on, I came to a realization. I realized that when non-vergence depth clues weren’t complementary that those elements or perceptions in conflict required suppression to continue viewing without some sort of physical effect occurring (typically unpleasant such as headache, nausea, etc.).
My paper is a start to the investigation of the importance of supporting perception cues as it relates to stereovision.
*Vergence is the simultaneous movement of both eyes in opposite directions to obtain fixation and the ability to see depth.
*Accommodation is the automatic adjustment in the focal length of the lens of the eye to permit retinal focus of images of objects at varying distances. It is achieved through the action of the ciliary muscles that change the shape of the lens of the eye.
Filed under 3D, 3D HDR, 3D Health Issues, 3D Motion Picture, 3D Photography, autostereoscopic, binocular disparity, binocular rivalry, HDR, High Dynamic Range, Perception Conflicts, S3D, stereopsis, stereovision
Last week I was in Washington, D.C. for a meeting with the director and exhibit staff at the Smithsonian Museum of Natural History. While I was there, I decided to pull out the 12 camera rig and take some architectural photographs on the mall, since the weather was reasonably nice.
I specialize in “life size” photography and people have asked me how I can take a life size photograph of a building?
The answer is that for me, the “life size” definition is the size that is perceived at the point of capture. So, if the cameras are several hundred yards away from the subject, life size would be the size as perceived viewing from that distance, not the actual size of the building. My goal with each photograph is to present it as you would see the imagery in real life.
The most impressive photograph I was able to take was from the end of the mall at the Lincoln Memorial looking towards the Washington Monument with the Capital building in the background and the reflecting pool in the foreground. There were just enough clouds passing by so that I was able to get perfect highlights with just the right amount of specularity on the Capital building in the background. But in order for the shot to represent true to life luminance detail I ended up needing to bracket the shot with multiple exposures (HDR photography). Normally, I am able to expand the dynamic range from the single shot, but in this sunlit outdoor example there was just too much range to capture even with my jeeped cameras.
Fortunately, not much moves when shooting buildings ;^) and I was able to have, in effect, what was a 1/2 second exposure as the cameras can fire three photographs in 1/2 second. Now the tough part comes, where I have to match and align 36 photographs and HDR merge them and interlace them into a single autostereoscopic image. Guess what? I’m still working on it ;^) But I will post an animation here in a few days to show the perspectives.
When I’m finished, I’ll have the most realistic photograph from the perspective of the Lincoln Memorial looking towards the Capital building that has ever been taken. That in itself would be worth the trip – add to that the meeting at the Smithsonian and this was a fantastic trip!
High dynamic range or HDR photography extends the number of luminance values in a photograph. What this means in layman’s terms is that dark parts of an image still have detail and don’t turn into a black blob of ink and bright parts of an image don’t blast to white. This is essential for 3D multiperspective photography because the most compelling photographs present to the eyes in the same way real life is presented to the eyes. Our eyes adapt instantly to changing light conditions and we see an amazing tonal range that extends far beyond a regular photograph with standard dynamic range. We see many subtle shades of darkness and white sand in the bright sun still has sand grain detail that is easily visible although we might have to squint. The dynamic range of our eyes greatly exceed the capability of cameras – both film and digital.
Take a look at this photograph as an example. Note on the right side standard dynamic range photo how the black fabric turns to black and the hair ribbon blasts to white. A considerable amount of detail is lost making it difficult and even impossible for some parts of the image to have clearly defined perspectives. (See blue circle blow up for detail)
One might argue that it is better for flat single perspective photographs as well. And for some that is true, however, often times a photographer wants to simplify the photograph or add dramatic lighting and these subtle changes in luminance (brightness) are less important. Also, with a flat photograph more detail in the background can conflict with the main area of interest in the photograph. Without dimensional depth to set it out, too much detail can be undesirable for a flat single perspective photo. With the example shown here one could argue that the detail of the black fabric takes away slightly from the baby. But in 3D there is a world of difference that has to be seen to be appreciated. That extra dynamic range provides the detail in 3D to clearly position the baby in it’s space and the stereovision perception is greatly enhanced. It looks far more realistic.
So, how do you get more dynamic range out of a camera? In my case, I have imaging sensors and special processing that extend the image data captured to 18 bits per color (red, green and blue)**. A regular camera with jpeg output is limited to 8 bits per color. Those extra bits I am able to obtain with my custom rig contain subtle changes in luminance levels that can be processed and printed to appear similar to the way the eyes would see them in real life.
Another way (actually the way most people do it) to make an HDR photograph is to take 3 or more photographs in quick succession with a bracketing camera option where the shutter speed is different for each photograph. Then using special HDR software they combine the different exposures into a single photograph with a broad tonal range. Sometimes the effects of this processing is very effective. For example, room interiors with windows look much more natural. Like anything though, it can be overdone and create very unnatural looking images. A big problem with this approach is that you are limited to shots that have no movement during the multiple exposures. In order for me to take action shots, I had to create a system that captures all of the data at once, at shutter speeds in the hundreths of a second range.
** A regular Canon Camera RAW image has 16 bits per color. I do use Canon sensors but I have been able to tweek things to get an extra couple of bits per color at the expense of error correction, which I must perform in a separate process with a computer. It will remain a secret how I do it, unless and until I am able to obtain a patent for the process. If you really have sharp eyes, you might also be able to detect that I’ve reduced color fringing and ringing around sharp luminance transitions. That’s another benefit to extra bits.