The First Photomosaic
In 2002, archaeologists excavating the properties in Regio VI, Insula 1 in the northwest corner of the city wanted to document the facade of all of the structures along via Consolare with a single image for study and project publications. However, the city block is over 110 meters long and the street is only four to six meters wide. It is impossible to stand back far enough from the structures either to view the complete insula or for it to be photographed as a single image.
The inspiration for the first photomosaic came from another imaging technique, panoramic photography. Panoramas are wide in comparison to their height and show a field of view greater than one can see, even up to 360˚. Panoramic photographs are created by recording an image (or multiple images) as a camera is rotated around a fixed vertical axis. Unfortunately, panoramic photography would not produce an acceptable image of the fronts of the buildings of Insula VI, 1. The narrow width of the street prevented the camera from viewing the city block from a far enough distance to eliminate perspective distortion. A panorama could record the entire street, but the facade would appear “fat” in the center, and “thin” at the ends. Also, the structures farther away from the camera would be seen at increasingly oblique angles. The intuitive solution was for the camera to be moved along the street rather than rotated around a fixed position.
Film was still the preferred photographic imaging media in the early 2000’s. It was determined that it was not feasible to combine multiple color slides or negatives into a single image. Multiple color prints could be spliced and glued together, but the resulting images would have shown a number of cut marks, could not be easily reproduced and the color quality would vary between the constituent pictures. The then emerging digital photographic technology was investigated to determine if it could produce acceptable results. A Sony digital camera was used for preliminary experiments to capture and combine sequential overlapping photographs. The results were promising.
The first photomosaic of Insula VI, 1 was created in 2002 utilizing digital technology to record and combine multiple photographs taken along the length of the street. The Sony Digital Mavica camera was employed for the photography and each picture was taken in landscape format (wide rather than tall). The camera was hand-held, and images were taken approximately every two meters to assure that each view would overlap the next. At the places where the buildings were taller than could be recorded on a single image, an “upper” picture was taken by tilting the camera.
The following summarizes the metadata for the photographs that comprised the first photomosaic of Insula VI, 1:
- Date: July 2002
- Camera: Sony Digital Mavica MVC-FD88 (1.3 megapixels)
- Image pixel dimensions: 640 X 480
- Image size at 72 pixels per inch (28.34 pixels per cm): 8.9” X 6.7” (22.6 cm X 17.0 cm)
- Photo file size: 120 KB
- Image format: Landscape
- Image Type: JPEG
- Focal length: Zoom set to 4.74 mm (35 mm film camera equivalent of 41 mm focal length)[1]
- Number of images: 123
The photographs were merged with Photoshop image editing software during the fall of 2002. A blank Photoshop base layer was created with a resolution of 72 pixels per inch (28.34 pixels per cm). Photographs were sequentially opened, copied and pasted onto the base as separate layers. Each new photo was placed so that it overlapped the previous one. The opacity of the latest image was reduced to approximately 50% so that the previous photograph underneath could be seen. The new image was then moved to match any details such as windows, doors, cracks, holes, etc. with the one below for correct placement on the background.
It became quickly apparent that the photographs were not the same scale. They each had been taken as far away from the opposing buildings as possible, and because the street varied in width, the distance from the camera to the facades varied. Each photo was, therefore, examined to see if it matched the previous one in size (scale) as well as position. Those that did not match were enlarged or reduced using the Transform/Scale tool in Photoshop that allows images to be proportionally altered. If the camera had not been level when the photo was taken, the Transform/Rotate tool was used to straighten the picture. In addition, photographs taken with the camera tilted upward to capture the tops of the taller buildings contained perspective distortion.[2] These images were corrected with the Transform/Perspective and Transform/Distort tools.
After a photograph was satisfactorily sized and placed with the adjacent image, the unwanted parts of both were removed using the Eraser tool, and the opacity returned to 100%. Overlaps were blended and adjustments were made to the color and contrast of the newest image to match all those that had been previously incorporated. The careful combination of individual photographs into a final image called to mind the mosaic floors seen throughout Pompeii. We therefore coined the term “photomosaic” to describe the finished work.
The following summarizes the metadata for the first photomosaic:
- Date: Fall 2002, improvements in Spring 2003
- Computer: Apple Power Mac G4, System 9 operating system
- Image editing software: Photoshop Version 6.0
- Number of combined photographs: 123
- Number of Photoshop layers: 125
- Background layer pixel dimensions: 13,600 X 2,160
- Background layer resolution: 72 pixels per inch (28.34 pixels per cm)
- Background layer dimensions at 100%: 188.9” (479.8 cm) X 29.9” (76.2 cm)
- Photomosaic image file size: 106.4 MB
At full size and at a resolution of 72 pixels per inch (ppi), a print of the finished photomosaic would have been nearly 16 feet (5 meters) long. The printed size would not only be very large, but also the resolution would be below the standard for minimum print quality of 300 ppi. The image resolution was therefore changed to 300 ppi, which decreased the length to a little over three feet (one meter). The image was then flattened (all layers combined) to reduce the final file size.[3]
The following summarizes the metadata for the flattened photomosaic:
- Flattened image pixel dimensions: 12,663 X 2,704
- Flattened image resolution: 300 pixels per inch (118.1 pixels per cm)
- Flattened image dimensions at 100%: 42.2” (107.2 cm) X 9.0” (22.9 cm)
- Flattened image file size: 26.4 MB
Even though rudimentary equipment had been used and the techniques were newly developed, the first photomosaic of Insula VI, 1 was unique and proved to be useful for study and documentation. The following advantages were recognized:
- A view of the structures was presented that is impossible to see on site.
- One image displayed a subject over 100 meters in length.
- The structures were recorded in color with reasonable detail.
- The facades of the buildings were all approximately at the same scale.
- The perspective views through the doors and windows of the structures gave a three dimensional feel to the image and added interest.
Although this first trial was considered a success, several shortcomings were identified in the image and with the recording and production techniques:
- Several errors were found that indicated that some of the photos had been improperly sized or placed.
- The image was not created to a known scale, and therefore objects could not be accurately measured.
- There was some doubt about the authenticity of the colors.
It was recognized that the future utility of the photomosaics would be dependent on their accuracy. The deficiencies were studied to determine possible corrective actions. In several areas, the wall surfaces of the structures did not have enough specific detail to assure that the photos were joined correctly. In another part of the city block, the individual ashlars in a long row all looked much alike. Their similarity again resulted in placement errors. A method of independent verification was required to assure that the photographs were precisely merged. Photogrammetry was identified as a technology that could possibly improve the photomosaic methodology.
[1] CCD sensors used in most digital cameras have a smaller surface area than that of a 35 mm negative. As a result, the focal length of a lens used on a digital camera will be different (usually larger) that it would be if used on a 35 mm film camera. The “35 mm equivalent focal length” of a lens is computed by multiplying the focal length of a lens by the multiplier factor that is published with the specifications of the camera.