Kodak Photo CD Conclusions and Recommendations

Conclusions

Used carefully, Kodak's Photo CD system can give excellent results. You must understand the various pitfalls of the system, and ways to work around them. By finding software to deal with the Lost Highlights Problem, Photo CD becomes significantly more useful. Although Kodak tries to pass it off as a problem in other people's software, the problem really lies within their own.

Recommendations

I believe that Kodak's Photo CD system can be improved significantly and produce results that will satisfy far more customers by making the following changes:

The Scene Balancing Algorithm

As with most automatic exposure systems, Kodak's Scene Balancing Algorithm falls flat on its face in many cases. Kodak should switch to a simple "Auto-Levels" style system with conservative black and white point settings like .05% or smaller. With my Polaroid SprintScan 4000 and these settings, I consistently got results significantly better than Photo CD. I would recommend only doing this auto-levels operation on the union of the Red/Green/Blue channels, not the individual Red/Green/Blue channels to avoid injecting any color casts. It appears as if the current algorithm acts on the luminance of the image which explains why it is possible for the Red channel to be totally blown while the other channels are OK.

Of course, there might be a performance penalty here that I'm not aware of. Kodak's current system might do a limited form of the above system to keep the speed up. It may be that aging PIW hardware can only process so many points in a histogram while cranking through 4 images a minute. What's really annoying is that the newer CS1200 has the exact same problems.

SBA is partially responsible for the Lost Highlights Problem, and completely responsible for Photo CD's poor performance with Black and White Negative films.

The Highlight Compression LUT

This Lookup Table works on the assumption that Luminance values in the Photo CD data over a certain point represent highlight information. Unfortunately, this assumption rarely is correct since you can't reasonably expect a computer algorithm to know the difference between a flat white, and a highlight. It's obvious that Kodak or someone agrees with me on this since the Photo CD players don't use a Highlight Compression LUT (and as a result they look great). Other inconsistencies include the fact that PCDLIB32.DLL's Highlight Compression LUT is different from the one in Kodak's Photo CD File Format Plug-In for Photoshop.

The Highlight Compression LUT should be removed from the File Format Plug-In, and PCDLIB32.DLL. This in combination with an improvement in the Scene Balancing Algorithm will result in a scanning system that is unbeatable.

The right way to handle PhotoCD images for professionals is to offer a conversion directly from PhotoYCC to Lab color, preserving the the full range of "Y" values as "L" values. This way the image can be edited in its full glory within any image editor that supports Lab (e.g. Photoshop). With Lab color, the photographer has full control over how the information encoded in the PhotoCD image is used. This is how it should be. Had Kodak done this from the start, PhotoCD might still be around as a professional format.

The "Source Profile" Field

In the Photo CD File Format Plug-In for Photoshop, the Source Profile field should be removed from this plug-in's dialog. It is totally absurd and causes nothing but confusion and bad color. The color space of a Photo CD file is an extended version of sRGB. It is not different for each different kind of film scanned. Whoever wrote this piece of software was very confused, or wasn't being told the whole story by the engineers at Kodak.

<- Back to my Photo CD page.