| Re: Any suggestions for a B&W colour scheme for publication images? [message #55936] |
Tue, 25 September 2007 08:30 |
Steve Eddins
Messages: 9
Registered: July 2001
|
Junior Member |
|
|
hradilv wrote:
> On Sep 25, 9:25 am, Tyler <hayes.ty...@gmail.com> wrote:
>> Hello All:
>>
>> I am in the process of submitting some figures to a journal for
>> publication (it has been accepted). I intend to save my supervisor
>> cash by submitting B&W figures for print, and keep the colour images
>> for the electronic version.
>>
>> Here is my problem. Originally, I've been using one of the EOS colour
>> schemes and the figures turn out great. There is excellent contrast
>> between regions, and the positive/negative values are clearly
>> distinguished. Sadly, switching the colour scheme to B&W tends to blur
>> these crucial distinct regions.
>>
>> [snip]
>
> Here's a reference: J. McNames, "An effective color scale for
> simultaneous color and gray-scale publications," IEEE Signal
> Processing Magazine, Vol. 23, No. 1, January 2006, pp. 82-87. (pdf:
> http://bsp.pdx.edu/Publications/2006/SPM_McNames.pdf)
I had a bit of correspondence with McNames about this article. It turns
out that his formulas assume a relationship between RGB and gray scale
that isn't quite accurate enough to get a fully monotonic scale in gray.
An alternative idea to plot a path through L*a*b* space, as I described
here:
http://blogs.mathworks.com/steve/2006/05/09/a-lab-based-unif orm-color-scale/
Since I don't know IDL, I can't provide a translation :-), but if you
have a way to convert between sRGB and L*a*b*, the rest of the math is
straightforward.
--
Steve Eddins
http://blogs.mathworks.com/steve
|
|
|
|
|
|
| Re: Any suggestions for a B&W colour scheme for publication images? [message #55938 is a reply to message #55937] |
Tue, 25 September 2007 08:05  |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
David Fanning writes:
> Good suggestions. If this doesn't work, I'd also try converting
> your RGB color vectors to HSV vectors with COLOR_CONVERT, then
> stretching your saturation vector as far as it will go, before
> converting back and loading into your color table. That's been
> known to work before, too, although sometimes the saturation
> is already stretched to the max anyway. (It was in the couple
> of color tables I just tried.)
I suppose you could try doing various log and power transformations
on the saturation vector. Here are a list of programs you can
find on my web page that might come in handy:
GMASCL
LOGSCL
ASINCSCL
CLIPSCL
GAUSSSCL
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.dfanning.com/
Sepore ma de ni thui. ("Perhaps thou speakest truth.")
|
|
|
|
| Re: Any suggestions for a B&W colour scheme for publication images? [message #55939 is a reply to message #55938] |
Tue, 25 September 2007 08:03 |
Vince Hradil
Messages: 574
Registered: December 1999
|
Senior Member |
|
|
On Sep 25, 10:01 am, hradilv <hrad...@yahoo.com> wrote:
> On Sep 25, 9:25 am, Tyler <hayes.ty...@gmail.com> wrote:
>
>
>
>> Hello All:
>
>> I am in the process of submitting some figures to a journal for
>> publication (it has been accepted). I intend to save my supervisor
>> cash by submitting B&W figures for print, and keep the colour images
>> for the electronic version.
>
>> Here is my problem. Originally, I've been using one of the EOS colour
>> schemes and the figures turn out great. There is excellent contrast
>> between regions, and the positive/negative values are clearly
>> distinguished. Sadly, switching the colour scheme to B&W tends to blur
>> these crucial distinct regions.
>
>> I have tried several values for STRETCH. Perhaps I was not using the
>> STRETCH command properly. Does anyone have any suggestions for values
>> of STRETCH or GAMMA_CT that have worked for their own publications in
>> the past? This problem is only further compounded by the fact that
>> XLOADCT seems to not want to load from my IDL script file, so I'm
>> limited to what I can do from the within the script.
>
>> For what it's worth, I have several figures to convert to B&W, all of
>> which have different maximums and minimums. Can I use that knowledge
>> to control STRETCH/GAMMA_CT with these values?
>
>> I have two example PNG files I can send to anyone if they would like
>> to see what I am talking about. Just email me with "GAMMA_CT" in the
>> subject heading.
>
>> Any ideas are greatly appreciated.
>
>> Cheers,
>
>> t.
>
> Here's a reference: J. McNames, "An effective color scale for
> simultaneous color and gray-scale publications," IEEE Signal
> Processing Magazine, Vol. 23, No. 1, January 2006, pp. 82-87. (pdf:http://bsp.pdx.edu/Publications/2006/SPM_McNames.pdf)
>
> I translated the matlab code here:
>
> function hyperbola, nc, ymaxa=ymaxa
>
> if n_elements(ymaxa) eq 0 then ymaxa=0.95
>
> a = sqrt( (1.0-ymaxa)^2 / (1.0-(1.0-ymaxa)^2) )
> xvals = findgen(nc)
> xs = 2.0 * xvals/float(nc-1) - 1.0 ; scale from -1 to 1
> nx = n_elements(xvals)
>
> y = 1.0 - sqrt(xs^2+a^2)/sqrt(1.0+a^2)
> y = y>0.0
>
> return, y
> end
>
> function triangle, nc
>
> result = findgen(nc/2)/float(nc/2-1)
> result = [result,reverse(result)]
>
> return, result
> end
>
> function colorspiral, nc, np, palname, smin=smin, twosided=twosided
>
> if n_elements(smin) eq 0 then smin=0
>
> ;nc=128
> ;np=2
>
> wn = sqrt(3.0/8.0) * hyperbola(nc)
> a12 = asin(1.0/sqrt(3.0))
> a23 = !pi/4.0
>
> if keyword_set(twosided) then begin
> t = sqrt(3.0)*findgen(nc/2)/float(nc/2-1)
> t = [-reverse(t),t]
> wn = [reverse(wn),wn]
> endif else begin
> t = sqrt(3.0)*findgen(nc)/float(nc-1)
> endelse
>
> r0 = abs(t)
> g0 = wn * cos((t-sqrt(3.0)/2.0)*np*2.0*!pi/sqrt(3.0))
> b0 = wn * sin((t-sqrt(3.0)/2.0)*np*2.0*!pi/sqrt(3.0))
>
> pm, 3, 3
> plot, r0, g0
> plot, r0, b0
> plot, g0, b0
>
> rgpol = cv_coord(from_rect=transpose([[r0],[g0]]),/to_polar)
> rg1 = cv_coord(from_polar=[rgpol[0,*]+a12,rgpol[1,*]],/to_rect)
> r1 = reform(rg1[0,*])
> g1 = reform(rg1[1,*])
> b1 = b0
>
> plot, r1, g1
> plot, r1, b1
> plot, g1, b1
>
> rbpol = cv_coord(from_rect=transpose([[r1],[b1]]),/to_polar)
> rb2 = cv_coord(from_polar=[rbpol[0,*]+a23,rbpol[1,*]],/to_rect)
> r2 = reform(rb2[0,*])
> g2 = g1
> b2 = reform(rb2[1,*])
>
> plot, r2, g2
> plot, r2, b2
> plot, g2, b2
>
> spiral = bytscl([[r2],[g2],[b2]])
>
> ;smin = 0
> smax = nc-smin-1
>
> spiral = congrid(spiral[smin:smax,*],nc,3)
>
> openw, lun, palname+'.pal', /get_lun
> printf, lun, palname
> for i=nc-1, 0, -1 do printf, lun, string(format='("#",3z2.2)',spiral[i,
> 0],spiral[i,1],spiral[i,2])
> free_lun, lun
>
> return, spiral
> end
Oops, forgot to mention that you can remove the plot commands
(diagnostics) and also that I write out the palette for use with
AFNI. But you get the idea.
|
|
|
|
| Re: Any suggestions for a B&W colour scheme for publication images? [message #55940 is a reply to message #55939] |
Tue, 25 September 2007 08:01 |
Vince Hradil
Messages: 574
Registered: December 1999
|
Senior Member |
|
|
On Sep 25, 9:25 am, Tyler <hayes.ty...@gmail.com> wrote:
> Hello All:
>
> I am in the process of submitting some figures to a journal for
> publication (it has been accepted). I intend to save my supervisor
> cash by submitting B&W figures for print, and keep the colour images
> for the electronic version.
>
> Here is my problem. Originally, I've been using one of the EOS colour
> schemes and the figures turn out great. There is excellent contrast
> between regions, and the positive/negative values are clearly
> distinguished. Sadly, switching the colour scheme to B&W tends to blur
> these crucial distinct regions.
>
> I have tried several values for STRETCH. Perhaps I was not using the
> STRETCH command properly. Does anyone have any suggestions for values
> of STRETCH or GAMMA_CT that have worked for their own publications in
> the past? This problem is only further compounded by the fact that
> XLOADCT seems to not want to load from my IDL script file, so I'm
> limited to what I can do from the within the script.
>
> For what it's worth, I have several figures to convert to B&W, all of
> which have different maximums and minimums. Can I use that knowledge
> to control STRETCH/GAMMA_CT with these values?
>
> I have two example PNG files I can send to anyone if they would like
> to see what I am talking about. Just email me with "GAMMA_CT" in the
> subject heading.
>
> Any ideas are greatly appreciated.
>
> Cheers,
>
> t.
Here's a reference: J. McNames, "An effective color scale for
simultaneous color and gray-scale publications," IEEE Signal
Processing Magazine, Vol. 23, No. 1, January 2006, pp. 82-87. (pdf:
http://bsp.pdx.edu/Publications/2006/SPM_McNames.pdf)
I translated the matlab code here:
function hyperbola, nc, ymaxa=ymaxa
if n_elements(ymaxa) eq 0 then ymaxa=0.95
a = sqrt( (1.0-ymaxa)^2 / (1.0-(1.0-ymaxa)^2) )
xvals = findgen(nc)
xs = 2.0 * xvals/float(nc-1) - 1.0 ; scale from -1 to 1
nx = n_elements(xvals)
y = 1.0 - sqrt(xs^2+a^2)/sqrt(1.0+a^2)
y = y>0.0
return, y
end
function triangle, nc
result = findgen(nc/2)/float(nc/2-1)
result = [result,reverse(result)]
return, result
end
function colorspiral, nc, np, palname, smin=smin, twosided=twosided
if n_elements(smin) eq 0 then smin=0
;nc=128
;np=2
wn = sqrt(3.0/8.0) * hyperbola(nc)
a12 = asin(1.0/sqrt(3.0))
a23 = !pi/4.0
if keyword_set(twosided) then begin
t = sqrt(3.0)*findgen(nc/2)/float(nc/2-1)
t = [-reverse(t),t]
wn = [reverse(wn),wn]
endif else begin
t = sqrt(3.0)*findgen(nc)/float(nc-1)
endelse
r0 = abs(t)
g0 = wn * cos((t-sqrt(3.0)/2.0)*np*2.0*!pi/sqrt(3.0))
b0 = wn * sin((t-sqrt(3.0)/2.0)*np*2.0*!pi/sqrt(3.0))
pm, 3, 3
plot, r0, g0
plot, r0, b0
plot, g0, b0
rgpol = cv_coord(from_rect=transpose([[r0],[g0]]),/to_polar)
rg1 = cv_coord(from_polar=[rgpol[0,*]+a12,rgpol[1,*]],/to_rect)
r1 = reform(rg1[0,*])
g1 = reform(rg1[1,*])
b1 = b0
plot, r1, g1
plot, r1, b1
plot, g1, b1
rbpol = cv_coord(from_rect=transpose([[r1],[b1]]),/to_polar)
rb2 = cv_coord(from_polar=[rbpol[0,*]+a23,rbpol[1,*]],/to_rect)
r2 = reform(rb2[0,*])
g2 = g1
b2 = reform(rb2[1,*])
plot, r2, g2
plot, r2, b2
plot, g2, b2
spiral = bytscl([[r2],[g2],[b2]])
;smin = 0
smax = nc-smin-1
spiral = congrid(spiral[smin:smax,*],nc,3)
openw, lun, palname+'.pal', /get_lun
printf, lun, palname
for i=nc-1, 0, -1 do printf, lun, string(format='("#",3z2.2)',spiral[i,
0],spiral[i,1],spiral[i,2])
free_lun, lun
return, spiral
end
|
|
|
|
| Re: Any suggestions for a B&W colour scheme for publication images? [message #55941 is a reply to message #55940] |
Tue, 25 September 2007 07:57 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
R.G. Stockwell writes:
> I do the same thing. Usually I can just use the default B&W scheme
> (loadct,0) and natively create the image that way. Perhaps reducing
> the number of contour or shade levels will help the contrast. Or perhaps
> adding a contour overlay on top of the image may help.
>
> One trick that sometimes works (but usually doesn't) is to
> use your normal color scheme (EOS) but not set the /color keyword to
> the ps device command. That seems to directly translate your color scheme
> into B&W. The problem, is that distinct colors can get mapped to similar
> shades of grey.
Good suggestions. If this doesn't work, I'd also try converting
your RGB color vectors to HSV vectors with COLOR_CONVERT, then
stretching your saturation vector as far as it will go, before
converting back and loading into your color table. That's been
known to work before, too, although sometimes the saturation
is already stretched to the max anyway. (It was in the couple
of color tables I just tried.)
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.dfanning.com/
Sepore ma de ni thui. ("Perhaps thou speakest truth.")
|
|
|
|
| Re: Any suggestions for a B&W colour scheme for publication images? [message #55942 is a reply to message #55941] |
Tue, 25 September 2007 07:45 |
R.G.Stockwell
Messages: 163
Registered: October 2004
|
Senior Member |
|
|
"Tyler" <hayes.tyler@gmail.com> wrote in message
news:1190730302.624006.228950@d55g2000hsg.googlegroups.com.. .
> Hello All:
>
> I am in the process of submitting some figures to a journal for
> publication (it has been accepted). I intend to save my supervisor
> cash by submitting B&W figures for print, and keep the colour images
> for the electronic version.
I do the same thing. Usually I can just use the default B&W scheme
(loadct,0) and natively create the image that way. Perhaps reducing
the number of contour or shade levels will help the contrast. Or perhaps
adding a contour overlay on top of the image may help.
One trick that sometimes works (but usually doesn't) is to
use your normal color scheme (EOS) but not set the /color keyword to
the ps device command. That seems to directly translate your color scheme
into B&W. The problem, is that distinct colors can get mapped to similar
shades of grey.
Cheers,
bob
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
