AGOCG logo
Graphics Multimedia VR Visualisation Contents
Training Reports Workshops Briefings Index
This report is also available as an Acrobat file.
  Back Next Title

Conclusions - Recommendations and Guidelines

Recommendations: but which is best?

Many factors will have a bearing on the answer to this question, some of these are:
  • the image source(s) you are using - for instance the camera used provides RGB and composite video outputs but not S-video, so it is not possible to comment on how the cards which can handle this input would perform. RGB signals give better quality images but this option increases the cost of the cards.
  • your current system may not have the space or clearance for a (or another) full length card (note the height of the Targa card). If you have no free slots, no spare computer or a laptop then your choice becomes not much of a choice at all be aware that there are other parallel port (and some PCMCIA) image capture devices in addition to the Video Snapshot and since their installation is likely to be very simple you should find it easy to evaluate them before purchasing.
  • is it to be single function or, for example, will it mainly be used for motion video capture and still image capture be very much a secondary function (e.g. VideoBlaster RT300)
  • and, of course, cost. You might decide the quality of the images produced by the Targa card is what you want but it is not cheap and the extra funds may not be available.
From the devices available for this evaluation it was considered that Truvision's Targa card (2000) gave the best overall image quality even at 16 bit colour. The quality is attributed to the RGB input available on the card. A range of image capture, handling and output software is available (specifically) for this card. In this instance TIPSPLUS was used but it is not very 'user-friendly', especially when compared to current software user interfaces. Among the less expensive devices the Black Widow Video Snapshot (199) performed well, its major drawback being the loss of shadow detail (though if you are lucky enough to have a variable output on your video source you may be able to compensate for this). It also has the advantage of being instantly transferable between computers (although this portability may be seen as a disadvantage as the device may be easily 'lost'), as long as they have a suitable TWAIN aware application but do not only have high speed parallel ports. For those who need to capture images when visiting other locations this is probably the sort of device to use.

Guidelines and tips for Image Capture and other considerations

  • Cards that use pass-through/hardware to overlay the incoming video image may have input bandwidth limitations, which may mean that you have to reduce the refresh rate and/or resolution of your display card. If you are then using the same computer for image editing as well as image capture you may find yourself swapping cables and changing software setups on a regular basis. Capture cards which use software overlay have an advantage in that driver updates are easily installed.
  • RGB inputs gave much better results than composite compare the images captured using the TARGA and Kingfisher cards from their RGB inputs to those obtained via their composite inputs, then compare both these sets to other cards (all composite).
  • Although the Targa captured images at 16 bits (in our set up) they appear much 'cleaner' and sharper than the 24 bit images from the Kingfisher, although the effects restricted by the dynamic range of 16 bit capture could sometimes be seen when looking at areas of gradual tonal change (dithering; slight granular appearance). Note: the Targa saves its images as 24 bit even when captured at 16 bit.
  • No matter how good (or bad) the quality of the captured image, it can be very easily degraded by poor handling/processing (much like photographs themselves). With this in mind here are a few guidelines that have been found useful in image capture :
    • capture at the highest colour depth
    • capture at the highest resolution; for video sources the number of pixels vertically should preferably be equal to the number of scan lines - any more or less will, almost inevitably, imply some form of averaging/interpolation.
    • clean system - stray dust tends to obey Murphy's law and be on the most important (and hardest to re-touch) part of the image.
    • for subsequent image processing : have the appropriate hardware and software tools and set them up properly. Editing 24 bit images on a 256 colour display is not a good idea. Setting up image editing software to match the gamma of your monitor and your room lighting can drastically affect the appearance of the displayed image - and don't forget to set up the monitor too.
    • for images which are (or destined to be) greyscale, either :
      - capture as greyscale, although this may restrict the corrections available if the source image is colour rather than monochrome.
      - or capture as 24 bit colour then convert to greyscale; avoid using 256 colour (even as an intermediary format). If captured as 8 bit colour and then converted to greyscale continuous (greyscale) tones are not as smooth, as some of the grey tones have been lost when the 256 colour palette was created. Ringing, seen as colour fringes at 24 bit, will (almost) disappear on conversion to greyscale as the colour, which stands out against the predominately grey background, is converted to grey.
    • high frequency lines/edges (e.g. fence posts, railings) which are perpendicular to the video scan lines tend to cause 'ringing', especially on composite inputs. If the slide has such lines and they are predominantly vertical, a better image may be obtained by rotating the slide through 90 - and then rotating the image back again in your capture application before saving.
    • Now that you have the images protect them :
      - write protect and/or make a working copy of the original file(s) it is very easy (especially towards the end of a long editing session) to overwrite the original by forgetting to change the file name or format.
      - save work often; some image editors have a limited undo capability (often just the last operation) and you may have to start over if you make an unrecoverable error.
    • The following sequence of operations has been useful when reducing the size and/or colour depth of images (as is likely for most CBL material). Note: this leaves out less common operations such as sharpening/blurring, background removal etc.
      • crop borders, but do not resample; this reduces filespace/RAM requirements which will speed up subsequent operations
      • adjust contrast/brightness
      • adjust colour balance/saturation
      • repeat the above, in small steps, as needed. Too great a jump in settings at each stage may mean that detail is irretrievably lost, e.g. a large increase in image brightness will lose highlight detail and subsequent colour adjustments are likely to be working only on white in those highlights.
      • now that the image is as 'right' as you can get it this is the time for resizing.
      • palette reduction; if this was done earlier (even before resizing) other operations are likely to change the colour 'population' of the image, which would compromise the palette.
      • SAVE and BACK UP

Back Next Title Home