More digital resolution...

Paul R Montague (mailto:Montague-Paul@OPHTHALMOLOGY-PO.OPHTH.UIOWA.EDU)
Mon, 22 Aug 1994 09:49:00 CST

Message-Id: <mailto:199408221520.KAA00292@library.wustl.edu>
Date:         Mon, 22 Aug 1994 09:49:00 CST
From: Paul R Montague <mailto:Montague-Paul@OPHTHALMOLOGY-PO.OPHTH.UIOWA.EDU>
Subject:      More digital resolution...
To: Multiple recipients of list IMAGELIB <mailto:IMAGELIB@ARIZVM1.BITNET>

     Jim Linder writes...

"Subject: Resolution, Pixel color depth, Archival Stability of Magnetic Media

"I strongly agree with Ray Lauzzana about questioning resolutions under 1024x1536 as being good enough for accurate representation. An extremely important element to consider is the color bit depth for each pixel..."

Thanks to Jim and Ray for valuable thoughts. I think both are correct...

Ray is right on the money with his assessment of required depth. When we have full color images digitized at 8 bits per color plane, stepping is sometimes obvious, especially when rendering large monochrome fields with gradually changing density (sky, for example). At 12 bits per color, stepping is greatly reduced. In a black and white image, eight bits give us 256 levels of gray while 12 bits give 4,096 levels. For color images, 24 bits (8 bits for each of three color planes) gives 16.8 million colors while 36 bits give 68.7 billion colors.

The point regarding degradation of magnetic media over time is an excellent one. One characteristic of digital images stored on magnetic media which we might consider, is that a copy of the digital image file is exactly as good as the original file. In fact, a 50th generation copy is exactly as good as the original, which is absolutely untrue for film. Perhaps a policy could be implemented that would require images to be copied periodically, avoiding the media degradation

As for the 1024x1536 issue, any pixel resolution below that is noticeable to most observers. We don't have responses for our study yet. All we know so far is that it is highly improbable that anything below 1024x1536 will match the resolving power of 24x36mm film. It is equally improbable that 6000x9000 pixels at 16 bits per color is required. Somewhere in the middle the truth must reside. I'm afraid I didn't make my point very well: Film images and digital images are different. While mathematical models can be used for comparison of resolution, the differences in the media might prevent those comparisons from being accurate.

Thanks for the input. These are things I'm trying to work out too, and it help to bounce the thoughts off others with similar interests...

mailto:paul-montague@uiowa.edu