| Re: Re IDL w/ 12-bit grayscale? [message #48238] |
Tue, 04 April 2006 20:33 |
Andrew Cool
Messages: 219
Registered: January 1996
|
Senior Member |
|
|
If Klaus Ertel is ever diagnosed with a brain tumour, I do hope that
he'll insist that his MRI
data is dumbed down to no more than, say 10 grey levels, because his
radiologist
can't discern anymore than that. ;-)
Andrew
|
|
|
|
|
|
|
|
| Re: Re IDL w/ 12-bit grayscale? [message #48268 is a reply to message #48267] |
Mon, 03 April 2006 18:14  |
Andrew Cool
Messages: 219
Registered: January 1996
|
Senior Member |
|
|
As someone who's lived with a brain tumour for 10 years now, I can say
that I'm more than happy for the True Believers to worship 4096 levels.
Nothing can be overkill when it's your brain they're looking at... ;-)
Andrew
|
|
|
|
| Re: Re IDL w/ 12-bit grayscale? [message #48269 is a reply to message #48268] |
Mon, 03 April 2006 16:19 |
JD Smith
Messages: 850
Registered: December 1999
|
Senior Member |
|
|
On Mon, 03 Apr 2006 09:41:05 -0700, kuyper wrote:
>
> Ertel, KG (Klaus) wrote:
>> Hello Mike,
>>
>> Is it really worth the hassle? My guess would be that you can't
>> distinguish more than, say, 10 grey levels by eye. ...
>
> Try this:
>
> image = rebin(replicate(1,256)#indgen(256),1024,1024,/sample)
> tvscl,image/16
> tvscl,image/8
> tvscl,image/4
> tvscl,image/2
>
> I could still easily detect the jumps in brightness level with image/4.
> It was a little bit harder for image/2, but I think I could still see
> them.
This is a really bad test. The eye is exquisitely sensitive to
brightness contrast, but very insensitive to absolute fluxes. In
fact, in the world of panorama creation (which I dabble in), blending
stitched images together can leave unsightly seams in the sky when the
actual bit value on two sides of a seam differs by only 1 (in 8-bit
per channel images)! Have a read:
http://enblend.sourceforge.net/banding.htm
This usually motivates people to go to 48bit images, to leave room for
post-processing. Yes, you heard me correctly.
Here's a better test:
IDL> a=bytarr(512,512)+128b & for i=0,511,64 do a[i+32:i+63,*]+=i/64+1b & tv,a
when do you see the bars separate? The deltas are a single (1/256)
jump in gray.
JD
|
|
|
|
| Re: Re IDL w/ 12-bit grayscale? [message #48275 is a reply to message #48269] |
Mon, 03 April 2006 09:53 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
kuyper@wizard.net writes:
> I agree - 4096 grey levels is overkill.
Gentlemen, I think you miss the fact that there are
True Believers (e.g., radiologists) among us. They don't
*care* if they can't see it, they believe it is important.
End of discussion. :-)
Cheers,
David
P.S. Let's just say it's not bad business to
believe in what True Believers believe in.
Especially if you sell expensive hardware.
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.dfanning.com/
|
|
|
|
| Re: Re IDL w/ 12-bit grayscale? [message #48276 is a reply to message #48275] |
Mon, 03 April 2006 09:41 |
James Kuyper
Messages: 425
Registered: March 2000
|
Senior Member |
|
|
Ertel, KG (Klaus) wrote:
> Hello Mike,
>
> Is it really worth the hassle? My guess would be that you can't
> distinguish more than, say, 10 grey levels by eye. ...
Try this:
image = rebin(replicate(1,256)#indgen(256),1024,1024,/sample)
tvscl,image/16
tvscl,image/8
tvscl,image/4
tvscl,image/2
I could still easily detect the jumps in brightness level with image/4.
It was a little bit harder for image/2, but I think I could still see
them.
> ... 12 bits would give
> you 4096 grey levels, slight overkill. I very much prefer false colour
> representations anyway.
I agree - 4096 grey levels is overkill.
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
