Demonstration Quality Scanning Digital Camera

Andrew Davidhazy 
Imaging and Photographic Technology Department 
School of Photographic Arts and Sciences 
Rochester Institute of Technology

For many years now I have been trying in vain to obtain a sophisticated line-scan digital camera back or a simple linear CCD array camera and although I was close in a couple of instances, things never panned out quite right.

In a technology teaching environment having access to the basic operating systems of devices is very important but destroying sophisticated instruments that I did manage to borrow was out of the question. So, for many years I drew parallels between digital scanning imaging devices and their "photographic" predecessors but could never actually easily demonstrate the connections between them.

Well, I have finally achieved my "digital dreams" by purchasing a KYE hand document scanner. Before long the device was reduced to its basic components and soon thereafter was doing things its designers probably never foresaw as possibilities.

In case you did not realize it yet, this scanner is a device that can be rolled over flat originals and as it is moved it sequentially but continuously captures an image of whatever passes underneath the scanning head where a fluorescent light fixture illuminates the subject.

The scanner is hooked up to a PC by way of a high speed data transfer card and it is designed to make scans that cover an area of up to 4 inches wide by about 14 inches in length. The software that is provided with it allows a user to make multiple passes over a subject wider than 4 inches and to join them together into a larger reproduction.

As the scanner is rolled over the subject the area illuminated by the lamp essentially moves past a CCD array consisting of many light sensors arranged in line fashion. These detect and record changing image information from the subject that appears to move past the array since the image of the subject is focused on the CCD by a lens focused on the print surface.

The CCD array is filtered in such a manner that it simultaneously records red, green and blue information for a full color reproduction of the original.

At the time I bought it the scanner cost less than $50 but it did need a computer to hook up to. The device I am using in conjunction with the scanner is a Dell PC running Windows 95.

Thinking that given the low cost of the scanner I would not be loosing much and after making a few scans of prints to familiarize myself with its operation, I decided to take a peek at the "guts" of the system and looked for the screws that fastened the top to the bottom of the scanner case. They were not difficult to find and soon the two were parted.

It turns out the cover was not really needed to run the scanner and after proving this I proceeded to identify the location of the CCD array. This was easily accessible and it was located at the end of a plastic "funnel" assembly equipped with a deflecting mirror in the chamber. In the middle of the funnel was a lens, the function of which was to reduce the image of the subject's width to the length of the CCD array. The use of a "folded" design made for a nice, low-profile, instrument.

I tried to make images by simply "panning" around my room with the scanner while it completed a scan cycle and I got some fuzzy but recognizable images of my lamp and chairs, the outside parking lot, etc. Promising!

Anyway, the CCD array was connected to the rest of the instrument by way of a multi pin connector. I unplugged this and then removed two screws that held the CCD array in place. The array could be cleanly removed from the imaging funnel and the lens was now easily visible within it.

The next thing was to try to see if the scanner would work with the sensor removed. It did. In fact, I also unplugged the connection to the fluorescent tube and the scanner worked without it also! The only thing it could not do was to complete a calibration process but this did not prevent the software from allowing the scanner to run through a scanning cycle.

At first glance the CCD array looked to be about 20mm long and encased in a glass enclosure about 40mm long, and the whole device attached to a copper clad PC board which in turn had the connecting socket leading to the rest of the scanner attached to it.

Disclaimer: Before proceeding further, I want to mention that the following material is based on a cursory bit of "reverse engineering", however, the principles that you will read about are, I hope, valid. The punctilious reader may find slight discrepancies in the measurements and reported operating characteristics of the device.

Now I taped the scanner's CCD array to the back of an old Canon Ftb camera whose shutter I had locked in the open or B position. The scanner still worked but this time the images recorded were a bit sharper and more recognizable than when I was simply "waving the scanner about"!.

The major problem was that the connecting wires of the scanner were much too short to work conveniently with the camera. So I made an "extension" from some multi-wire cable that I had lying around. I essentially added a female "D" socket near the CCD array and at the end of a multi-wire cable soldered to the connections on the PC board I attached a male "D" connector. This way the CCD array could be hooked up to to scanner or mot, to facilitate transportation.

While that "taped" version of the system demonstrated the soundness of the improvisation, eventually I decided to make the system a little more permanent and dedicated an old Olympus camera body to the purpose. I filed out the focal frame of the camera to the extent that the PC board on which the CCD array was built could be attached to the body in such a manner that the CCD array was approximately in the same location as the film would be in the focal plane of the camera. Since I installed the CCD array along the long dimension of the focal plane, the mirror mechanism of the camera was undisturbed and could still be used to focus the camera lens. A mark was drawn on the groundglass of the camera to indicate where the location of the CCD array would be aimed at once the mirror was lifted up to uncover the CCD array to light from the lens.

Even though the CCD array is epoxied in place, to this date I only have made a fairly temporary installation but that is all that is needed for now. In fact, making the installation permanent would make the array _less_ useful for teaching purposes.

Finally, since the scanner would not be rolling over a surface and the rolling wheels be controlling the rate at which data from the scanner's CCD got transferred to the computer, I took a DC gearhead motor and connected its shaft by way of a rubber band to the shaft of a slotted wheel designed to rotate between a photodetector and an IR light source relaying information to the scanner's software as to the data acquisition rate. Because the scanner itself never was moved, the DC motor therefore provided an artificial rolling rate parameter to the scanner's software which it would use to determine when the scanner had achieved it's preset scanning aspect ratio or dimension.

The data acquisition rate would then control the degree to which a subject appeared to be reproduced normally in terms of aspect ratio or whether it appeared compressed or stretched out in the preliminary record of whatever the scanner was "fed" in terms of input image.

In the case of contact scanning the aspect ratio was always properly maintained but once the advance mechanism was disconnected from the subject itself, as in the case of the decoupling the rolling motion from the scanning action, then the aspect ratio of the reproduction could be altered at will.


1. Panoramic Photography

The earliest application that I looked into was that of making a panoramic photograph. This is achieved by the simple expedient of placing the camera on a tripod with the CCD array arranged so that it is parallel to the tripod's axis of rotation (meaning in a vertical direction) and while the scanner is presumably rolling over a surface, scanning it, slowly rotating the tripod's column while the camera's shutter is locked in the open position. Using the array lined up along the long dimension of the 35mm camera's image gate and with the camera in the turned as if to make a vertical shot, the largest possible number of pixels or light sensors of the array are exposed to light from the camera lens.

Later measurements seem to indicate that the sensor is about 20mm worth of pixels. This means that the sensor could also have been placed along the short dimension of the 35mm camera's focal plane but th4e consequence would have been fairly major "surgery" of the camera's body casting and a probable destruction of the mirror movement.

Using a very short focal length lens, such as a 14mm lens, it is possible to make an almost full 360 degree image showing acceptable aspect ratios of objects included in the scene. The aspect ratio of any image or scene reproduced through these means is roughly a function of the vertical angle of view of the lens and the horizontal angle of view that one covers. If the camera is turned too quickly or the focal length of the lens used is too long the objects in the scene will appear much too narrow ... horizontally compressed.

When making panoramic records of scenes covering a full 360 degrees around the camera the rotation of the camera can be accomplished by simply turning the tripod's center post by hand. After doing this a few times I opted to place the camera on a motorized tripod head. This made the rotation very even and uniform.

The actual focal length that would be required to make a full 360 degree image with as little distortion as possible is hinted at by the 4x12 (or 1x3 aspect ratio of the scanner's controlling program) and the actual length of the CCD array. Since the active portion of the array is about 20 mm long (proven by making a "contact" image of a mm scale placed directly on top of the CCD array and allowing it to scan) this would make the length of the recording "surface" (if there was a surface!) about 60 mm long. Since the focal length required for any panorama is equal to the radius of a circle whose circumference is equal to the length that we want our photograph to cover, this would call for a 10 mm focal length lens. Admittedly this is rather short but generally one can tolerate some horizontal compression and/or one can always not quite cover 360 degrees!

One can easily figure out the number of "horizontal" degrees that any lens would cover (using this particular array/software combination) by taking the 60 mm and multiplying times 360 degrees and dividing this by the focal length of the lens in question multiplied by 2 and then by 3.14.

So, a 35 mm focal length lens would cover 60x360=22600 divided by 35x2x3.14=220 or 22600/220=103 degrees.

In any case, it is interesting to see how the image is captured in almost real time as the scanner successively displays the surrounding scene onto the computer screen!

Finally, it is not absolutely necessary to capture the image in an undistorted state. One can purposefully compress the horizontal angle of view so that it fits in the available digital image "length" and the use a "uncompressing" scheme in an image processing program (such as Photoshop) to stretch the horizontal dimension to an appropriate length or "throw away" height data to such a point that the number of pixel in a height dimension are appropriate for the number of pixels available in a horizontal direction.

2. Peripheral or "rollout" photography

The recording of the complete outer surface of (usually) cylindrical objects has been the specialty of a very few photographers since the technique of peripheral photography was first considered, developed and applied in the late 19th century in archeological museums where Greek and other vases from antiquity were photographed "in the round" to record the total outer circumference of designs added to their surfaces.

Shell Oil Company developed a camera in the 1930's designed to photograph pistons and cylinders to show areas of wear, etc. I started to experiment with peripheral photography in the mid 1960's making peripheral portraits of people. These portraits often did not look very much like the actual persons!

For peripheral photography I aimed the linear CCD array forward and I stood on a turntable in front of the camera after having pressed the "scan" command on the software controlling the scanner. The idea was that as I rotated in front of the camera the lens would project continuously changing features from my head onto the linear array located in the camera image plane.

The scanner would be functioning as a "non contact" printing press where my head would supply, over time, the changes in information that the scanning array would then store in the computer's memory. I tried to make a couple of turns or so during the time it took the scanner to believe it had scanned a 14 inch long print. Success!!

I also applied the camera to traditional peripheral photography by placing a vase on a turntable and adjusting the image size and the rotation rate so that a more or less acceptable reproduction of the surface features of the vase was displayed in the final image.

3. Linear strip photography ("photofinish" and moving array over subject applications)

The idea this time is to aim the array at some location in space and have a subject go across that location or take the camera as a whole and move it past a stationary subject.

This particular application of the system seems more limited than most because the data capture and transfer rate are relatively slow and thus any real application to photofinish photography is somewhat far fetched.

Nevertheless, if one keeps in mind that the objective is to move the image of whatever subject one aims the lens at (normally one would adjust the data capture/transfer rate) at such a speed that the image compiled by the scanner has the same aspect ratio of height to length as the original, then the system "works". Even if there is some slight discrepancy between the reproduction and the original that could amount to as much as 50% in either direction, the reproduction will usually be recognizable as being an image of a particular subject.

In the photograph shown here I walked back and forth across the line in space delineated by the scanning elements and each time I was walking forwards but sometimes from left to right and others from right to left. Note how the images seen in the final reproduction all show me heading in the same direction. You might want to spend a few minutes trying to figure out why this is so!

4. Scanning array 2D photography (moving array across film plane)

I also decided to try duplicating "focal plane scanning" as is done with sophisticated scanning camera backs such as some made by Phase 1 and by Finelight, and others and typically used in 4x5 cameras.

For this the camera is aimed at a particular scene and the linear array is slowly moved from one side of the camera's image gate to the other while the scanner accumulates the changing image information. In commercial cameras the linear array it moved across the image plane by sophisticated transport mechanisms at a rate that is predetermined by the software that controls the scanner.

In my case I simply moved the array by hand. (BTW, this was done before I fixed the array in place!)

I moved the array along the long dimension of the film gate as the scanner "assumed" it was rolling over a stationary subject.

The interesting aspect of this improvised image-plane scanner is that, because it is hand-driven, the speed can be varied at will or by accident. Ultimately, for the best quality reproduction, the time that it should take the array to move across the image plane is such that the aspect ratio of objects in the scene is reproduced properly in the final image.

Again, my first experiments were done by simply pulling the linear array slowly across the film plane, maintaining contact between the front surface of the array and the film positioning rails of the camera. If the resulting image looked too compressed this indicated that the next time I should move the array more slowly.

More precisely, though, the time that it should take the array to move the 36 mm distance can be determined from the fact that one needs to end up with an image that bears a relationship between its own "length" (in this case 36 mm) and the total image length the scanner "generates" during a full scan.

No extraordinary measures were taken to exclude light or infrared from entering the camera back, which was open, and reaching the CCD array. An infrared opaque cloth was simply draped over the back of the camera as the array was moved across the film plane.

Needless to say, in this mode the subject must remain still and the lighting must not change during the time the array is recording the image plane. However, some interesting distortions can be generated by varying the rate of motion of the array, changing the lighting on the scene or, indeed, placing the subject itself in motion.

If such changes do occur while the scanning process is taking place, then one can easily see an effect that in traditional photography is often talked about but seldom seen, namely focal plane shutter distortion!.

Finally, if one insists on using this scanner to generate "finished" work it is often quite possible to do a little bit of electronic retouching on the final image to hide those areas where image dropout is most evident.

An unexpected problem/benefit: Infrared sensitivity

Another "problem" that the designers of this scanner (and most other scanners as well) had to deal with was the inherent sensitivity of CCD devices to infrared. In order to accurately record only the red, green and blue information from the scanned subject most every manufacturer of scanners deals with this in a similar way. They light the subject with light that contains little infrared or they include an IR absorbing filter in front of the array. Usually, even if this filter is not totally effective in eliminating infrared, this is acceptable because all one needs to do is to make sure that enough of the infrared is removed so that the color response of the device is not noticeably adversely affected by the presence of residual infrared.

But once the array has been removed from the camera and is used to record everyday scenes, the array is bound to run across situations where there is an overabundance of infrared present. This results in unwanted color casts to the final image. To eliminate these as much as possible one is then forced to include an infrared absorbing filter someplace ahead of the scanner's picture elements. Usually this is accomplished by placing an IR absorbing (but light transmitting) filter in front of the camera lens.  Sometimes a cheap heat absorbing glass from an old slide projector will work adequately.

Or, one could take this deficiency and make it into an opportunity! The scanner can be used as a rudimentary infrared imager. Think about this, an infrared-capable _panoramic_ imager! To make a panoramic image one rotates the camera as described above. To make this device into an infrared imager one places an Infrared transmitting, but light absorbing filter, such as a visually opaque Wratten 87 or 88A or similar, over the lens. Or simply use this as a device for making interesting images only by infrared energy.

In fact, the CCD array is extremely sensitive to IR and it is quite simple to light a scene with ordinary tungsten illumination and covering the lens with an 87C filter and, as shown in the attached illustration, experience no problems at all in making infrared images. The attached illustration was made at one stop larger aperture then the full color image shown in the panoramic photography section. Yet it was made through a totally visually opaque 87C IR filter.

Hopefully this article gives you ideas for experimentation as well.