Creating 3D Video for use in Vision Training
I did the following research in tandem with the School of Optometry and the College of Education‘s Technology and Learning Center in the University of Missouri-Saint Louis. As an optometry student with a strong background in audio-visual technology, I was asked to creatively utilize video techniques in order to develop new methods for vision training. This particular project dealt with developing a certain form of 3-dimensional video that is different from the customary polarized method which did not provide enough light, a particular problem found in amblyopic (lazy eye) patients. We are seeking to determine whether or not stereoscopic (3D) video will be useful as part of vision training for patients who suppress an eye, which is often the case for patients with amblyopia. There have been many approaches for treatment, including patching the dominant eye and cycloplegia. Our approach is to recreate stereoscopic conditions in such a way that they can be manipulated and monitored more specifically than is inherent in other methods. We currently have a working model that may ultimately push the amblyopic eye to function, while still allowing the dominant eye to gather visual cues.
- Two perspectives of scenes or objects are simultaneously filmed with digital video cameras (Picture A), which serve to mimic the points of view of the two eyes that are normally involved in vision
- These two video images are then presented to the individual in such a way that they are “fused” together into a single 3-dimensional construct
- The patient is asked to sit in front of a monitor or a screen (Picture B). He or she views two video images through prism glasses or a mirror device that will superimpose the two images (Picture C)
- The dominant eye is presented with an image that has either reduced contrast or has been modified with blur
- Periodically, the patient is monitored in order to determine whether or not they are achieving fusion
Protocol for Producing useful Video
- The digital video footage was filmed with a camera lens pupillary distance of 9cm.
- Although High Definition (HD) cameras were used, current technologies do not allow for burning HD discs. Therefore, camera settings were adjusted to the maximum resolution for Digital Video (DV).
- Several things are kept in mind while shooting the footage: a. The cameras should be kept level with one another (a level was integrated into the camera set-up). b. Cameras often have auto-luminance mechanisms. These were left active. However, if one camera fell in the shade while the other did not, this would result in an imbalanced end result. c. The camera operator needs to include some kind of cue, which would later allow the video to be edited together exactly in time. d. The scene should have objects at varying distances and be interesting to potential patients.
- A general video editor (such as Apple’s iMovie) is used to edit scenes that are to be used and also to organize what will become the Left and Right channels. These scenes were exported at full quality as DV (.dv) files.
- These files were imported to a more advanced video editing program (Apple’s Final Cut Pro). The project frame size was set to 1450 by 480 pixels.
- The Left video channel was placed with its center coordinates at -364 and the Right video channel at +364.
- Despite best efforts to keep the cameras level, vertical displacement between the two images may still exist. When this is the case, vertical coordinates can be adjusted evenly between the R and L channels in order to balance this out. (When doing so, the Frame size should be reduced from 480 in order to crop off uneven image edges.)
- Integrate Shapes and place them at -370 and +370. These were created in Adobe’s Photoshop with image sizes of 720 x 480. They must be exported in a format that allows for transparency (e.g. TIFF) and is compatible with your video editing software.
- Export using the following settings (set each sequence setting similarly) • Frame Size Custom 1450 x 450 • Compressor H264 (set at maximum quality) • Key Frame Rate 24 • Multi-Pass Encoding
- This results in a video file that can be either viewed in Quicktime on a computer, or loaded into a DVD software program for burning a disc compatible for all DVD players.
Note: There was a large amount of trial and error involved in developing this protocol. There are many options available, some of which will take away or improve image quality, file size, and so forth.
It appears that this process may lead to a viable form of vision training. The original design was to reduce the contrast in one channel; however, the Gaussian blur is an attractive alternative. In subjects with normal stereopsis, it is possible to pick up stereoscopic cues from the blurred eye. This often occurs while the other eye (no blur) is integrating the details into the fused construct. The optimal result for patients requiring vision training would be to adapt the patient in steps, presenting variable amounts of blur, to the dominant eye. Eventually, the patient would learn to use the eye that is being suppressed, begin to fuse the images with blur in the dominant eye, and then begin to gather stereoscopic cues. As a result of this work, it appears that the technology exists which makes the design attractive stereoscopic video possible along with a regimen of vision training. There are many ideas and issues that have emerged during the course of this project that were left unanswered. The most significant issue is that it is difficult to monitor whether or not the patient is achieving true fusion and/or stereopsis. There are many monocular depth cues available in the video along with motion, which complicates the issue. Some options were explored as far as superimposing shapes with stereoscopic cues and fusion crosses; however, the best method is yet to be determined.
The fountain video was one of the first pieces of footage that I worked with. You will notice that two red ovals appear on each side; you may even be able to see that one of them is offset when you compare the two channels. When wearing prism glasses, one of the ovals will appear to be closer. This was one idea that I felt may be useful in helping to determine whether or not the patient is effectively using stereoscopic vision.
This video was filmed in a Digital Video format. While you may not be able to tell from this compressed streaming web version, the results on the final DVD product were of decent quality. It was filmed at the St. Louis Zoo. The green circles are there to help the patient fuse the two images. If the images are fused, the will see a red cross in the middle of the green circle. One of the channels is blurred in some of these scenes. We found that it was still possible to obtain disparity cues from the two images, despite significant blurring of one video channel.
If you are interested in following up on this research project, please contact Dr. Carl Bassi (Associate Professor and Director of Research and Graduate Studies for the College of Optometry) firstname.lastname@example.org
Trackback from your site.