I’ve been working with stereoscopy for a few years now but I’m still exploring what I would consider the very basics of what can be achieved with stereoscopic depth creatively. Playing with a stereoscopic camera is probably one of the fastest ways to explore different ideas and begin discovering what works and what doesn’t. Over the past couple of years I’ve been exploring the world of stereoscopy with a FujiFilm W1 and have been busy building a mental document about things that work, things that don’t and how to work around or take advantage of these things. Beyond the general principles of stereoscopy figuring out the techniques takes a lot of time and experimenting so below are some examples of what I’m exploring and what I have discovered along the way.
All af these images are embedded from 3DF33D so you can look at them in whatever 3D format suits you. I’m aware that most people don’t have a proper 3D setup but the classic cyan/red anaglyph glasses will work fine. You will need to set your 3D mode on 3DF33D and then refresh this page to see them all here in that mode.
The first thing you realise with a camera like the W1 is that you are limited to taking photos at a certain distance as the distance between the lenses are fixed. I think this is a good boundary to have. I have read many papers (back when I used to read such things) on creativity and cognition suggesting that boundaries can be increase creativity as you try and push beyond those boundaries or try and achieve something unique within them. So although you might not be able to get exciting stereoscopic images of huge landscapes, there’s still lots of playing to be had with all kinds of different environments, and of course the camera is really designed for happy snaps of the family etc.
At first it seems very difficult to get a good close up shot as most initial attempts have far too much distance between the minimum and maximum parrallax and are painfull to look at. To begin with I avoided closeups and kept around 2 meters away from most subjects, but once you control the distance between the closest and furthest objects, it is possible to achieve really good results that have a huge amount of depth within a very small distance. It is in fact this sort of shot that I find myself taking the most as the detail in the depth that is achievable makes the objects so tangible. The image below is a great example of how much depth can be achieved with a close up shot. If you could see anything beyond this tree stump, like a bit of the ground below it, it would be be painful to look at as the parallax would be too large.
I enjoy the tangibility of these kinds of images so much that I seem to have become obsessed with taking pictures of walls. They make up around 30% of all photos from a 3D shoot. Leaves are another obsession in depth. I love leaves.
This image is teetering on the edge of the pain threshold when looking at it fullscreen on a 27 inch monitor as there is so much depth between the top of the shoe and the ground, but just looking around the shoe at the ground is an adventure in depth.
I think it’s really important to remain aware of traditional picture making and cinematic techniques and test them out in 3D to see if they still work as expected or if they have alternative effects, good or bad. The classic cinematic vertigo shot can be enhanced with a little stereoscopic depth. The images below don’t really work in 2D as there aren’t enough depth cues in the rest of the image for your mind to build the depth required, they appear very flat and wouldn’t really have any physical effect on the viewer. When viewed in 3D, they create an odd sensation of vertigo even though the distance is relatively short, but this kind of sensation would normally require a huge abyss of space with other depth cues to have the required effect. This technique may not be that useful when it comes to the huge, sweeping, cliff hanging, vertigo shots, I just need to find a precipice to give it a go.
I’m surprised there haven’t been many 3D horrors appearing at the cinemas as I think many stereoscopic techniques would work really well to heighten suspense and distort reality. I haven’t played much with these ideas using the camera, the image below is looking at the roll angle of the camera to test if the sense of balance is enhanced with the addition of stereoscopic information and whether this can be used, as it is in many horrors, to create a sense of uneasiness. I don’t think the image is enough to make any conclusions.
One issue I found when colour grading the image was having to keep the mid-tones of the window so as not to lose the depth information on the fence spikes. If I pushed the contrast more, which I wanted to do, the spikes would have been lost in the background and the depth information would have been lost. This probably wouldn’t have been an issue if this was a video and the camera was moving but it still highlights the importance of lighting a shot and creating a good amount of depth using tradition 2D techniques. A rim light on the spikes would aid the problem, but generally it would benefit from a full lighting rig and probably some fog.
You can see in the image below, I’ve also added a scratched film overlay onto the back wall, following the same depth angle. It’s interesting to compare it to the image above.
It’s also interesting to compare the difference in percieved contrast in the 2D image and the 3D image. It seems that stereoscopic images don’t really need contrast in the same way that 2D images do. Sometimes in 2D, contrast is used to separate objects in space and emphasise lit edges. In 3D, the objects are separated by definable space, which might reduce the need to increase image contrast. There are many other theories in this area, these are by no means conclusions.
I really want to play with the distortion of depth across the image more, to see if it can produce some interesting effects or whether it just causes brain shear. I saw a really interesting video where subliminal frames of people had been comped into only one of the stereoscopic channels, which triggered all kinds of ideas for breaking a viewers mind. The eye is usually drawn to things that don’t look right, and is normally something to avoid as it can break the attention of the viewer from what they should be looking at, but this technique has been used to great effect in cinema to purposely grab the attention of the viewer to something, they wouldn’t ordinarily look at. Changing the colour on an object in one stereo channel would give it an odd sheen, so would immediately focus the attention of the viewer.
Another area of play is adding affects in post or manipulating the images and even playing with the depth within the image. The image below is one of several tests with creating insets of different depth, but as a side effect it demonstrates the inherent issues with using the left or right channel for the 2d version of an image or video. If you view the 3d version of the image the two inset circles appear centred to one another and visually balanced. When the the left or right image is viewed on its own, the circles are slightly offset and feel awkwardly balanced. To solve this problem a further centre image would need to be used. I’m not suggesting anybody would do that for any but the most extreme cases, it just highlights the issue well.
Below is an example of comping other stuff into the images in post. By using the 3D camera and some stereoscopic scripts, it is possible to add other 3D elements into the photographs. I use StereoPhoto Maker to extract the left and right channels from the .mpo files and also for some alignment and colour correction. I then use Pinkau’s stereoscopic scripts for After Effects to create stereoscopic comps to apply any grading, effects or overlays to both channels with ease. The above image was created by pasting an arrow object from Adobe Illustrator onto the mask of a 3D solid in AE, then duplicating the solid and rotating and scaling it in 3D space. To make sure it fits into the depth of the photo, I have a locked comp window of the anaglyph comp open on a second monitor to refer to. It’s only really a quick play, but it shows how easy it is to insert basic 3D elements into a photo you have taken to heighten the sense of depth or do something creative.
The below image is a big mix of interesting problems and ideas. The photo was taken in a restaurant around a meter away from my face, and the background was several meters behind me, so the image is quite painful to look at.
I have highlighted the maximum positive parallax on the anaglyph image below. It is not impossible to converge the image, it is less than our inter ocular distance, however, it it’s a big step in depth from the background to face. This requires a very quick and strong accommodation of the eyes to move between the background and the foreground, which can be quite painful if not tiring.
To overcome this problem I have created an individual mask in After Effects for both the left and the right images, once again using the features of the stereoscopic scripts to apply these quickly. This simply removes the background from the image and any problems it was causing, as my idea for the image works better without the background anyway. This is a really quick mask, with a large feather in it, with more time this could be tightened up around the head to fit exactly.
Beyond this, it’s a case of playing with ideas and experimentation. Adding a couple of 2D overlays at different depths is something I have been playing with recently. They add a certain look to the image as well as creating an interesting volume of depth.
I also added some impact graphical arrows so they align stereoscopically with the side of the face:
Below is a screen grab of the After Effects setup for the 3D arrow overlay that comes out of the mouth towards the camera. The camera seperation is then converged using a positional offset in the master left comp to place the initial part of the arrow in the mouth. You can see the stereo comp in anaglyph on the left and the important values have been exposed on the timeline.
And finally some grading:
Here’s the stereoscopic version again for comparison:
Another idea that I have been experimenting with is creating windows within an image like a split screen using frames to enclose the spaces behind. The image below is created from several stereoscopic images placed beneath a frame on the zero plane. The images all have a varied depth budgets or appear on different planes so they create a strange effect when juxtaposed next to each other. It’s much easier to see the subtler gradients of depth when it’s compared to another image with a different angle of depth. The shoe appears miniaturised against the stones, although I’m not really sure why, as the interaxial separation is fixed… Maybe this has something to do with the post convergence and scaling of the images? Nevertheless, it’s a very interesting image to me because of this. All the photos are taken from around the building where I live in Leicester.
Below are some ideas I have been working on for a short film. These images are playing with some of the previously mentioned ideas of layering multiple 2D layers in 3D space and creating windows of varying depth. These images were created without any photos at all, they are all composited elements placed at varying depths to produce the 3D image.
The image below is another whistle on the creative pipe, but it shows how you don’t need everything to have true stereoscopic depth for our minds to build the 3D image. The mountain is a 2D image but it’s very hard to convince yourself that it isn’t 3D. The brain is very good at understanding other depth cues and building 3 dimensional spaces without any stereoscopic information. All of these other depth cues, such as scale, perspective, parallax, saturation and dof can produce a very strong sense of depth and have been perfected over centuries of 2D image making. Our eyes are also too close together to see much, if any stereoscopic information in very large or distant structures such as mountains. This is why the stars appear to be the same distance away. I don’t think there is anything wrong with using hyperstereo to achieve greater stereoscopic depth in larger scenes, but it is always good to remember how our perception of depth works and how that affects the image I’m making.
The final image is another stereoscopic concept image for a short film I’m working on. It’s just as much an initial test using the VRay render engine in 3ds Max (I usually use Mental Ray or Final Render) as a creative experiment. I think the parts of the particle overlay that intersect with some of the other imagery are a little confusing to look at and will probably be even more so when there is movement. It’s interesting to see how some of the ideas that have been attained from my stereoscopic adventures with a 3D camera can translate into CG based creative work.