Scanner Comparison: PhotoCD v Sprintscan

by Philip Perkins

This is a comparison of scans from the same negative on two scanning devices:

• Kodak PhotoCD
• Polaroid Sprintscan 35 Plus

Here is a description of the scans:

PhotoCD:  I opened the original PhotoCD image in Photoshop at the maximum resolution of 2048 x 3072 pixels. In Photoshop, I then cropped the image to include just the galaxy. I then adjusted the Levels for correct colour balance (black level) and image brightness. Only the middle sliders were adjusted, so no contrast adjustment was done. I also checked for correct black level with the Eyedropper tool set to sample average of 5 x 5 pixels.

Sprintscan:  The scanner used was a Polaroid Sprintscan 35 Plus.  The same negative as above was scanned at maximum resolution of 2700 dpi. I cropped the image in the prescan software to match the PhotoCD image. I made a very slight contrast adjustment in the prescan software to try to match the contrast of the PhotoCD image. Then I carefully adjusted the exposure controls in the prescan software so that the black level was correct and the brightness closely matched the brightness of the PhotoCD image. After importing the image into Photoshop I checked the black level with the Eyedropper tool set to sample average of 5 x 5 pixels. I did not adjust or crop the image in Photoshop.

Note: the images below have not been integrated into the background. They may be saved individually for inspection in Photoshop if required.

To produce the images for Figure 1 I took each scan and rescaled it to exactly 35% of its original size. The relative scaling and aspect ratio of each image was preserved. The rescaling was only to render the images in a size suitable for web page display.

To produce the images for Figure 2 I took each original scan and cropped a closely matching area from each image. No rescaling was done. The purpose of Figure 2 is to enable a comparative inspection of the images at the original scanned image scale.

The following differences may be observed:

Image Scale:  From the above images it is obvious that the scanned resolution is different. I do not know the exact resolution of the PhotoCD image but I believe that it is around 2200 dpi. I did not try to match this with Sprintscan. My approach was simply to compare images based on the highest available resolution from each scanner. Hence the Sprintscan image was scanned at 2700 dpi. However there may be a further difference: does the vertical linearity look slightly different?  Is the Sprintscan image slightly more elongated or is it my imagination?

Image Noise:  By comparing the images in Figure 2 it is apparent that the Sprintscan scan has more noise in the image. Originally I believed that this was due to inherent noise within the scanner since there did not seem to be any other reasonable explanation. Subsequently Chuck Vaughn demonstrated that this could not be due to scanner noise since it did not appear when scanning a neutral density filter. Subsequently I conducted a similar test as part of the APML Scanner Test. This demonstrated conclusively that the 'noise' I was seeing was in fact film graininess. But then why does the PhotoCD scan show much less noise? I don't really know, but part of the answer may be that because of the lower resolution of the PhotoCD scan, it is effectively undersampling the grain pattern, hence it does not show up so much. However I am not sure that this is the complete answer - it may be that Kodak use some form of grain reduction as part of the PhotoCD process.

Colour Resolution:  By comparing the images in Figure 2 it is apparent that more colours are visible in the PhotoCD image. This is especially noticeable by comparing the centre of NGC 5195 (the companion galaxy below M51). There are subtle yellows and greens in the PhotoCD image which are not visible in the Sprintscan image. In the spiral arms of M51, more blues and reds are visible in the PhotoCD image.

On the basis of these observations I believe that PhotoCD offers a high level of quality that may even be higher than the Sprintscan for critical work. This is despite having to make significant post-scan adjustments to the PhotoCD image. On the right you will see the original PhotoCD image, rescaled but otherwise unadjusted. You will notice that both the brightness and colour balance are all wrong. Some fairly significant adjustments had to be made to the levels in Photoshop. Despite this, the PhotoCD image displays higher quality and less image noise.

Does this mean that the Sprintscan is a poor scanner?  No, I don't think so at all. In fact in my own tests it significantly outperformed the Nikon Coolscan LS-1000. The Sprintscan had less artifacts and image noise than the Coolscan, and displayed a wider range of subtle colours.  I think the answer is simply that the PhotoCD scanner is in a different league. It is a professional quality scanner that probably costs in the region of $20,000. If so, then the PhotoCD scanner is ten times more expensive, and it certainly should be better. Equally this comparison demonstrates just how well the $2,000 Sprintscan stands up against PhotoCD.

This is a fairly harsh comparison. It is not usual to compare the output of products with a tenfold price difference. Normally  one would seldom need to display or print images at 100% of maximum scanned resolution. For all practical purposes the output of the Sprintscan is of very high quality and the image noise does not become apparent unless fairly aggressive enhancement is needed of a single underexposed (astronomical) negative. But for critical work it may be interesting to note the quality difference that may be obtained with PhotoCD.

Note: I have no connection with Kodak or PhotoCD. In fact I will never use the PhotoCD service again because they substantially damaged my negatives. In the UK they use scanner operators who have no idea how to handle valuable photographic material. On the other hand I have invested my own money in the Sprintscan. Therefore I hope that this is a fair comparison.