Tim Slattery
01/08/2001 12:43 PM
To: James VanEe <jiv2@cornell.edu>
cc:
Subject: Re: Misc.: Scanners vs. Digital Cameras for Capturing Gel Images
(Document link not converted)
James VanEe <jiv2@cornell.edu> on 01/04/2001 11:00:57 AM
To: Recipients of ABRF List <abrf@aecom.yu.edu>
cc: (bcc: Tim Slattery/RM/USR/SHG)
Subject: Re: Misc.: Scanners vs. Digital Cameras for Capturing Gel Images
>Hi. I'm not an expert at gel image capture or analysis, but I can
>think of reasons why a digital camera may not provide the results you
>need. A reasonable quality scanner will collect more information
>than a digital camera (I'm not sure of the upper limits in camera
>technology (4 megapixel/$2000 maybe?). Taking the low road, where
>pric/"performance" is comparable, a 2 MegaPixel camera will collect,
>well, 2 Megapixels of information. a scanner at 300 dpi scanning a 5
>inch by 5 inch gel will collect (300 X 5) X (300 X 5) = 2,250,000
>pixels (units of information). pretty comparable, but most scanners
>will scan at least 600 dpi and with the camera you will be contending
>a lot more with ambient light issues, getting the gel you want to
>digitize at the right length from the camera so you take full
>advantage of each "pixel" on the camera ccd. I think this could
>affect how well you can calibrate your analyses. For densiometric
>analysis, I'm pretty sure you want a transmissive image (scanner with
>transparency adapter) and not a reflective one (camera).
>
>If anyone can refute my lay assumptions, please do! I'm pretty sure
>there are commercial gel documentation systems using camreas and I
>may be incorrect that more information is necessarily better. perhaps
>I'm confusing documentation and quantitation.
James,
Since you asked, I'd like to correct a couple things. The importance of
resolution (dots per inch) is primarily dependent on what you are going to
do with the image for output, but little to do with the ability to detect
bands (up to the point where you are no longer getting 5-10 points across
the narrow dimension of the band). When you print your printer will output
300 dots per inch (these dots per inch are NOT the same dots per inch that
the printer manufacturers quote when they say, for example, the Epson has
1440 x 720 dpi printing, that is the ability to place the cyan, magenta,
yellow and black spots to create 1 "colored" spot). If you enlarge you
image to where there fewer than 300 dots per inch it will begin to look
blocky, as the individual pixels get larger. If you never need to print out
a picture larger than the typical snapshot a 2 megapixel camera is just
fine in terms of resolution.
For detection of bands dynamic range is far more important. In theory the
CCDs used in scanners and cameras are often the same if they come from the
same generation of electronics, and so should have the same dynamic ranges.
That is not how it works in practice however. Cameras are designed for
taking snapshots of the kids under average lighting conditions. This means
assumptions like; the overall lighting of a scene is 18% grey total
reflectivity. The camera then balances the exposure to fit those
assumptions. A gel on a lightbox (note this is transmitted light and
scanner or camera will almost always use transmitted light) is going to
have an average brightness much lighter than 18% grey. In compensating the
exposure the camera will often lose dynamic range data, so if you have a
faint band it may no longer be visible. Setting exposure manually can often
fix this, but it may not be easy or intuitive. In contrast, scanners make
far fewer assumptions about the object being scanned, and also make the
controls for setting range and exposure more accessible. The default
exposure from a scanner will probably give you a better image than the
default settings of a simple digital camera. Given a skilled operator and a
camera with full manual controls I would bet that the cameral will give a
better image (at equivalent resolutions) to a scanner with the same
generation CCDs. Lens flare (which is what I think you were getting at with
the "distance from the subject") is controllable, and there are far fewer
interfering surfaces in the camera A typical scanner has the object glass,
4 mirrors and the lens to traverse before getting to the CCD. In a camera
there is only the lens, and possibly one mirror. This should result in less
light loss and noise before getting to the CCD.
Of course, having said that, I should point out that there are scanners out
now in the $1500 range that use 14 bit per color CCDs. I have yet to see a
consumer digital camera with more than 12 bits per color. This gives you
two powers of two more colors to divide your range in to, making it easier
to find smaller changes. Not many software packages are able to use all 14
bits, so the scanners put out the "best 12 bits" of the 14 which will still
give you better data than a 12 bit scanner.
Despite my excess of detail, I agree with you. Given a walk up device, and
minimal operator training a scanner will probably give better data most of
the time. On the other hand if you just want a snapshot and are not worried
about picking out faint bands from the background a camera is faster, more
portable and more intuitive to use.
--tks
Tim Slattery
Berlex Biosciences
This archive was generated by hypermail 2b29 : Thu Feb 01 2001 - 12:44:47 EST