By Tamara Thomsen and Keith Meverden
Photo mosaic of the 123-foot schooner Rouse Simmons, which lies in 170 feet of water in Lake Michigan. The video was captured during one 25-minute dive, and the mosaic is assembled from 242 individual images and is 88,939 KB. Courtesy: Wisconsin Historical Society
sing photo mosaics to document underwater sites is not a new technique; it has been utilized by a generation of diving scientists. Advances in digital photography and digital image manipulation, however, have allowed more options for correcting image distortions, which in turn allows a greater freedom in image collection – a distinct advantage on deeper sites where bottom time is at a premium.

Traditionally, photo mosaic documentation began with a rigid frame or grid placed over the site to be documented, dividing the site into equal sections. A still camera, mounted atop a triangular tower, was then moved section by section, taking sequential photographs at a consistent distance and angle from the bottom. Keeping the camera at consistent depth and angle was of great importance, as variations in distance or angle created scale and proportioning errors that could not be easily corrected. This process required vast amounts of bottom time to set up the site and capture images.

Using digital video rather than a still camera maximizes productivity during limited bottom times, allowing the capture of literally hundreds of images in a matter of minutes while traveling over the site at a consistent depth and speed, freeing the diver from stopping every few feet to ensure adequate image overlap. Still images can later be captured from the digital video to create the photo mosaic.

The key to all methods of digital video capture is to keep the camera steady and at a perpendicular angle to the bottom. This requires a camera housing that is nearly neutral in buoyancy, as well as using good buoyancy techniques. To keep the camera at a perpendicular angle to the bottom, the authors utilize a simple plastic bubble level that is available at any hardware store. Attached to the back of the camera housing, the bubble level provides a quick visual reference for camera angle. These cheap plastic levels have been used effectively to depths of 175 feet without problems. To maintain a consistent depth, a small digital depth gauge can be affixed to the back of the camera housing, but a wrist-mounted depth gauge, rotated to allow viewing without moving the camera, has proven to be more than adequate.
Photo mosaic of the 75-foot scow schooner Ocean Wave, which lies in 110 feet of water in Lake Michigan. The video was captured during one 12-minute dive, and the mosaic is assembled from 80 individual images and is 63,714 KB. Note the diver off the port bow. Courtesy: Wisconsin Historical Society
To begin, the camera operator must first determine the optimum depth and lane width for the passes with the video camera. This is accomplished by holding the camera perpendicular to the bottom while viewing the site through the camera’s viewfinder. The goal is to create a balance between the captured image’s width and detail by changing camera depth above the site. While looking through the viewfinder, adjust the camera depth to achieve the widest possible coverage area while maintaining adequate detail within the viewfinder. Camera exposure often needs to be adjusted while determining the optimum depth. Obviously, sites with poor visibility will require the camera to be closer to the site, while sites with unlimited visibility will allow the camera to fly much higher above the site. If passes are made too far above the wreck, fine image details are lost. If passes are made too close to the wreck, the field of view will be so narrow that a very large number of images will be needed to construct the mosaic.

Once the optimum depth level has been determined, the coverage width of the camera lens will need to be determined by picking two objects on opposite sides of the viewfinder. The distance between these objects is the field of view, and all passes over the wreck should be approximately one half to two thirds this distance away from one another. This will allow adequate overlap between lanes to ensure the site is thoroughly captured with no “dead spots.”
With the optimum depth and lane distance determined, the camera operator is ready to begin “mowing the lawn” over the site. On small sites it is possible to simply swim the video camera over the site while maintaining a consistent depth, but this allows much more camera shake than other methods. Movement from diver kicks, adjustments to buoyancy, or even checking depth gauges creates small movements in the camera, which detract from the captured image quality.

The authors have found the most efficient method is to mount the camera to the nose of a DPV, allowing a diver to operate both the camera and DPV. The diver first finds the optimum depth and lane spacing as above; and, after turning the camera to record, simply flies the DPV over the site with the camera attached to the nose. Even with a long-bodied DPV, the bubble level mounted on the back of the camera is easily visible, and a wrist-mounted depth gauge can be easily read to maintain a consistent depth. This also allows one hand to remain free to signal other divers, equalize, or maintain buoyancy.

The DPV’s speed has a large effect on captured image quality. High speeds allow complete site coverage in a minimal amount of time, but create challenges for capturing overlapping still images from the video, and often result in blurred still images. For sites with large variations in relief or with fine detail, mosaic production is made easier by capturing the video in one direction only (i.e. all video passes are made from bow to stern, with the return pass un-filmed). This is especially helpful if there are many shadows created with artificial lighting or on bright, high-relief sites.

Once topside, the next step is to capture individual still images from the video. Start at the beginning of the first pass and begin capturing still images onto a large capacity memory stick using the photo capture feature on the video camera. Successive images should overlap by about one-third. Still images should be captured in as high resolution as possible, but it is not essential that the image quality be as high as that of still camera images. The final mosaic will be a compilation of many images and the completed mosaic will usually be quite large, making the resolution of each individual less important. All still images should be captured before assembling the mosaic. Move the captured images from the memory stick into folders labeled “Pass 1,” “Pass 2,” etc. Organization is key. All images should be reviewed as they are transferred from the memory stick. Occasionally a captured image will contain a lot of blurred movement, making it unusable. It is easier to recapture the image immediately than to wait and have to search the entire video for the proper location from which to re-capture the image.
Diver collecting video with nose-mounted Light&Motion camera on a Silent Submersion UV-26
Software such as Canon PhotoStitch that automatically assembles series of images to create panoramas is rarely successful in assembling underwater mosaics. The best results are achieved by hand-assembling the images in Adobe Photoshop. Begin by opening the images one at time in Photoshop. If the site has a large feature that can be used as a visual baseline, such as a ship’s keel, start the assembly process along this feature and assemble the mosaic outward from there. Once the first pass is completed, ensure the assemblage of images is straight and does not curve one way or another. If the first pass of assembled images curves, flatten the images (from the pull down menu under “Layer”) and then use the Liquify filter to straighten out the anomalies. The brush size and pressure may have to be adjusted. If the video was shot with a wide-angle lens, the best results are achieved by using only the middle portion of each image (due to edge distortion created by the wide angle lens). Extra time should be taken to ensure the image edges are accurately aligned. Some images may need to be transformed or scaled in order to account for slight changes in depth during the video capture. For more artistic control, the coloration and tone can be adjusted. If color matching of images is difficult, one option is to turn the completed mosaic into black and white.

Once assembled, a few things can be done to improve the final appearance. The most obvious detractor from a mosaic is visible edges of individual images, which can create a “patchwork” appearance. This can be reduced by flattening the image and cleaning the individual image edges with a combination of Photoshop’s Healing Brush, Patch, and Blur tools. Once the mosaic is flattened and the image edges are hidden, the mosaic can be scaled for accuracy. Significant site features should be measured during the dive, and these measurements can be used to adjust to mosaic to a specific scale. Photoshop’s Liquify filter is used to stretch the mosaic until the mosaic’s measured site features match the chosen scale. If only certain areas of the mosaic require stretching to scale, areas that do not need adjustments can be frozen while areas needing stretching are moved.

The important thing to remember is that some techniques will work better for some sites than others. The techniques described have worked for the authors on shipwreck sites upwards of 350 feet in length and to depths of 175 feet.

Tamara Thomsen manages the U.S. office for Delta P Technology, Ltd., and owns Diversions Scuba in Middleton, WI.

Keith Meverden works as an underwater archaeologist for the Wisconsin Historical Society’s Maritime Preservation and Archaeology Program, and owns Points North Diving.