function [hh,ScalingDataColorImg,ScaleColorDataClim,colorbarh] = bar4viacolor(varargin)
% bar4viacolor - 4-D bar graph : bar3 with another dimension via bar color.
% Two color schemes are available: The jet colormap and another that, like
% bar3, is a column group scheme whose groups colors vary mainly by saturation
% (see BarColorType below)
%
% function bar4viacolor(varargin)
% function [hh,ScalingDataColorImg,ScaleColorDataClim,colorbarh] = bar4viacolor(varargin)
%
% This function is a add-on hack on bar3 and so I've left the bar3
% documentation below. W/r to the bar3 documentation below, bar4 uses a
% Z sized matrix to specifiy the 4D color dimension. There are some
% options that are appropriate with the new facility but, as it so
% happpens, bar3 uses its own argument parsing function. So rather than
% hack that too, I extended the bar3 Z argument to be a struture s.t.
% one of the fields is Z and the rest the new 4D parameters. This "Z"
% structure goes in the first or second argument, just like Bar3, and all
% other Bar3 args stay the same.
%
% Because color is used as another dimension, an obligatory colorbar is
% automatically be displayed showing the range of colors.
%
% The Z structure consists of these fields:
% {.Z, .ColorScaleData, .ScalingOption, .BarColorType}
% (Note that ScalingOption and BarColorType have defaults and those fields
% do not have to be in the structure for their defaults to take
% effect)
%
% Z - 3D data just like bar3 where each element yields a bar.
%
% ColorScaleData - Same size as Z. Defines the 4D data.
% Note, negative values will yield bars with red edges. New if this
% argument is left out or empty than it will default to the Z data which
% then causes the bars to be Z colored.
%
% ScalingOption - {'a'} | 's' | 'c' | Clim - This controls how the
% Scaling Data is mapped to the colors:
% 'a' - autoscaling - if all ScalingData is between 0 and 1 nothing is
% done but otherwise the data is mapped to the entire color range.
% 's' - scaled - The ScalingData is mapped to the entire color range
% (like imagesc)
% 'c' - clipped - ScalingData between 0 and 1 is scaled to the entire
% color range. Data outside that is mapped to the first and last
% colors respectively (i.e. as if ScalingData was clipped to 0,1)
% Clim - [minvalue, maxvalue] - ScalingData in the range of minvalue
% to maxvalue is mapped to the first and last colors respectively
% (i.e. as if ScalingData was clipped to Clim like imagesc w/
% clim)
%
% BarColorType - 'c' | {'s'} - controls the color scheme used to color
% the bars.
% The default 's' option colors each bar by indexing into the jet
% colormap just like, say, an indexed image using imagesc
% The 'c' option is, like bar3, is a column group color scheme.
% Specifically, all the bars of a column are the same hue but vary by
% saturation, and to a lesser degree intensity, of each bar is varied to
% reflect the ColorScaleData.
% WARNING: *** concerning the use of 'c' options with figures containing
% multiple subplots/colorbars see *** below
%
% Additional output args: ScalingDataColorImg,ScaleColorDataClim,colorbarh.
% Only appropriate for the BarColorType=='c' option.
% ScalingDataColorImg - shows the range of colors available to this
% histogram. A RGB image with row being the hue used for the
% respective column of the histogram and the columns being
% the breakdown thereof.
% ScaleColorDataClim - is the ScalingDataClim implied or specified (and
% so is the appropriate item to specify the range of X labels for
% ScalingDataColorImg as per the image function)
% image(ScaleColorDataClim,1:size(ScalingDataColorImg,1),ScalingDataColorImg);
% colorbarh - handle to colorbar. For the 'c' option resizing this might
% be desirable but the colorbar won't auto resize or move nicely if you do
%
% Example:
%
% figure;
% Z = reshape(1:6*7,6,7); % 3D values, i.e. bar heights, increase in column order
% ScalingData= reshape(-20:21,7,6)'; % 4D values, i.e. bar colors, increase in row order
% bar4viacolor(struct('Z',Z,'ColorScaleData',ScalingData, ...
% 'ScalingOption',[-17,17],'BarColorType','c'));
%
% Here the Z data is just 1:42 reshaped to a 6,7 array and the
% ScalingData is a likewise sized array from -20:21 but whos sequence is
% in row major order.
% The construction of Z is such that each column will have bar
% heights greater than the previous column and the construction of
% ScalingData, along with the specification of BarColorType == 'c', will
% independly produce increasing saturation within each row. The negative
% ScalingData values (the first 20 boxes in row major order) yield red
% edged light bars. The Clim clipping value means that the ScalingData
% outside of -17,17 is as if it were -17,17.
%
% Here's the 's' color scheme w/ the same data
% figure;
% bar4viacolor(struct('Z',Z,'ColorScaleData',ScalingData, ...
% 'ScalingOption',[-17,17],'BarColorType','s'));
% Now the bar color are completely independent and correspond to the jet
% colormap
%
% Here's a 3d color version.
% bar4viacolor(struct('Z',Z,'ScalingOption','s'))
%
% Andrew Diamond
% EnVision Systems LLC
%
% WARNING: *** : Having color as a 4th dimension makes a colorbar obligatory.
% Unfortunately, colorbars seeem to be funny (in a black humor way)
% things. The 'c' option requires the use of absolute vs indexed
% colors but colormaps don't have that in mind. I hacked into the colorbar
% information to make the color bar an RGB image of the colors used but
% it doesn't seem to be 100% proper. In particular, the use of this type
% of colorbar seems to interfere with standard colobars in the same figure
% (through the use of subplots). So, if you are making a figure with
% multiple colorbars where any are not from bar4viacolor with the 'c'
% option, don't use the 'c' option. But, if all you have is a figure
% with one plot or all subplots use the 'c' option then I think its OK.
% At one point just making the colorbar an RGB image caused the fonts to
% go nuts (like a memory overwrite of some sort). Also, it would be nice
% to have a wider colorbar with the 'c' option but if you alter a
% colorbar's position (width) it stops being completely automatic and
% part of that is a fixed width for a narrow colobar. If anyone wants to
% fix this go right ahead. Please. I just wanted an RGB colorbar but
% instead I got the bloody Spanish Inquisition. The histogram code was
% trivial!
% -----------------------
% BAR3(Y,Z) draws the columns of the M-by-N matrix Z as vertical 3-D
% bars. The vector Y must be monotonically increasing or
% decreasing.
%
% BAR3(Z) uses the default value of Y=1:M. For vector inputs,
% BAR3(Y,Z) or BAR3(Z) draws LENGTH(Z) bars. The colors are set by
% the colormap.
%
% BAR3(Y,Z,WIDTH) or BAR3(Z,WIDTH) specifies the width of the
% bars. Values of WIDTH > 1, produce overlapped bars. The default
% value is WIDTH=0.8
%
% BAR3(...,'detached') produces the default detached bar chart.
% BAR3(...,'grouped') produces a grouped bar chart.
% BAR3(...,'stacked') produces a stacked bar chart.
% BAR3(...,LINESPEC) uses the line color specified (one of 'rgbymckw').
%
% BAR3(AX,...) plots into AX instead of GCA.
%
% H = BAR(...) returns a vector of surface handles in H.
%
% Example:
% subplot(1,2,1), bar3(peaks(5))
% subplot(1,2,2), bar3(rand(5),'stacked')
%
% See also BAR, BARH, and BAR3H.
% Mark W. Reichelt 8-24-93
% Revised by CMT 10-19-94, WSun 8-9-95
% Copyright 1984-2005 The MathWorks, Inc.
% $Revision: 1.30.6.5 $ $Date: 2005/04/28 19:55:59 $
ScalingDataColorImg=[];ScaleColorDataClim=[];colorbarh=[];
error(nargchk(1,inf,nargin,'struct'));
nin = length(varargin);
zeroscale=[]; %
% all saturation values (beteen 0 an 1) get raised to this power. Maximum
% intensity colors, at least on my monitor, look totally saturated after
% about 70% saturation. This pushed that a bit but it comes at the expense
% of more unsaturated looking low values.
% Should probably devise a color map by analytic formuala that lowers
% intensity significantly only for > 90% saturation which would then make
% high saturated values easier to discriminate
% Then I added the non linear intensity mangling to do the same thing so
% the need for this parameter has gone away.
SaturationPower = 1;
iZStruct=0;
for ivarargin = 1:min(2,length(varargin)) % The Z structure will be the 1st or second arg
argi = varargin{ivarargin};
if(isstruct(argi)) % first element in mag which is bars size and second is phase which is color adj
iZStruct=ivarargin;
defaultParams = struct('Z',[], 'ColorScaleData',[], 'ScalingOption','a', 'BarColorType','s');
argi = mergedefaultparams(argi, defaultParams);
if(isempty(argi.ColorScaleData))
argi.ColorScaleData = argi.Z;
end
if( length(size(argi.Z))~= 2 || length(size(argi.Z))~= length(size(argi.ColorScaleData)) || isempty(argi.Z))
error('Z and ColorScale Data must be the same sized non empty 2d matrices');
end
ScaleColorData = argi.ColorScaleData;
ScaleColorDataRaw = ScaleColorData;
varargin{ivarargin} = argi.Z; % bar3 machinery will not see the bar4 Z structure stuff.
if(~ischar(argi.ScalingOption) && ~(isnumeric(argi.ScalingOption) && all(size(argi.ScalingOption(:))==[2,1]))) % scaling info for color
error('scaling data type should be a of type character A)uto, S)caling, C)lipping or a 2x1 clim')
end
if(isnumeric(argi.ScalingOption))
ScaleColorDataClim = argi.ScalingOption;
if(ScaleColorDataClim(1) >= ScaleColorDataClim(2))
error('ScaleColorData clim(2) > clim(1)');
else
ScaleColorData = min(1,max(0,(ScaleColorData - ScaleColorDataClim(1)) ./ diff(ScaleColorDataClim)));
zeroscale = (0 - ScaleColorDataClim(1)) ./ diff(ScaleColorDataClim);
end
else
scalestr = argi.ScalingOption;
minScaleColorData = min(ScaleColorData(:));
maxScaleColorData = max(ScaleColorData(:));
switch(lower(scalestr(1)))
case 'a'
if(maxScaleColorData > 1 || minScaleColorData < 0)
ScaleColorData = (ScaleColorData - minScaleColorData) ./ (maxScaleColorData -minScaleColorData);
ScaleColorDataClim=[minScaleColorData,maxScaleColorData];
zeroscale = (0 - minScaleColorData) ./ (maxScaleColorData -minScaleColorData);
else
ScaleColorDataClim=[0,1];
end
case 's'
ScaleColorData = (ScaleColorData - minScaleColorData) ./ (maxScaleColorData -minScaleColorData);
ScaleColorDataClim=[minScaleColorData,maxScaleColorData];
zeroscale = (0 - minScaleColorData) ./ (maxScaleColorData -minScaleColorData);
case 'c'
ScaleColorData = min(1,max(0,ScaleColorData));
ScaleColorDataClim=[0,1];
otherwise
error('scaling data type should be a of type character A)uto, S)caling, C)lipping')
end
end
sepcolorsstr = argi.BarColorType;
switch(lower(sepcolorsstr(1)))
case 'c'
colgroupcolors=1;
case 's'
colgroupcolors=0;
otherwise
error('bar4 BarColorType, colgroupcolors, must be ''c'' for column group colors or ''n'' otherwise');
end
break;
end
end
if(iZStruct==0)
error(sprintf('%s requires a Z struct arg',mfilename));
end
[cax,args] = axescheck(varargin{:});
[msg,x,y,xx,yy,linetype,plottype,barwidth,zz] = makebars(args{:},'3');
if ~isempty(msg), error(msg); end %#ok
m = size(y,2);
% Create plot
cax = newplot(cax);
fig = ancestor(cax,'figure');
next = lower(get(cax,'NextPlot'));
hold_state = ishold(cax);
edgec = get(fig,'defaultaxesxcolor');
facec = 'flat';
h = [];
cc = ones(size(yy,1),4);
if ~isempty(linetype)
facec = linetype;
end
nrows=size(yy,2)/4;
if(~colgroupcolors) % each bar has its own 4d color
cmap = colormap('jet'); % scaling is w/r to saturation;
colormap(cmap);
for i=1:nrows % size(yy,2)/4
for k=1:size(yy,1)/6
cdatai = diff(ScaleColorDataClim) * ScaleColorData(k,i) + ScaleColorDataClim(1);
if(ScaleColorDataRaw(k,i) < 0)
edgei = [1,0,0];
else
edgei = edgec;
end
h = [h,surface('xdata',xx((k-1)*6+(1:6),:)+x(i),...
'ydata',yy((k-1)*6+(1:6),(i-1)*4+(1:4)), ...
'zdata',zz((k-1)*6+(1:6),(i-1)*4+(1:4)),...
'cdata', cdatai .* cc((k-1)*6+(1:6),:), ...
'FaceColor', facec,... %
'EdgeColor',edgei,...
'tag','bar3',...
'parent',cax)];
end
end
else % scaling data via saturation w/r to each column having the same hue
cmap = colormap('hsv'); % scaling is w/r to saturation;
orangedivider =find(cmap(:,1) == 1 & cmap(:,2) < 0.7 & cmap(:,3) ==0 );
orangedivider = orangedivider(end);
cmap = cmap([orangedivider:end],:); % chuck orange red start so end bars isn't too much just like start bars
colormap(cmap);
cmapScalingT = [1,1;size(yy,2)/4,1]\[0.5+0.001;size(cmap,1)+0.5-0.001];
% hsvbasecolori = rgb2hsv(basecolori);
for i=1:size(yy,2)/4 % each column (same hue)
basecolori = cmap(round([i,1]*cmapScalingT),:);
hsvbasecolori = rgb2hsv(basecolori);
for k=1:size(yy,1)/6 % each row
if(ScaleColorDataRaw(k,i) < 0)
edgei = [1,0,0];
else
edgei = edgec;
end
hsvbasecolori(2) = ScaleColorData(k,i) .^ SaturationPower;
hsvbasecolori(3) = SatToDecIntensity(hsvbasecolori(2));
colori = hsv2rgb(hsvbasecolori);
h = [h,surface('xdata',xx((k-1)*6+(1:6),:)+x(i),...
'ydata',yy((k-1)*6+(1:6),(i-1)*4+(1:4)), ...
'zdata',zz((k-1)*6+(1:6),(i-1)*4+(1:4)),...
'cdata',i .* cc((k-1)*6+(1:6),:), ...
'FaceColor', colori,... %
'EdgeColor', edgei,... % edgec,...
'tag','bar3',...
'parent',cax)];
end
end
end
if length(h)==1
set(cax,'clim',[1 2]);
end
if ~hold_state,
% Set ticks if less than 16 integers
if all(all(floor(y)==y)) && (size(y,1)<16)
set(cax,'ytick',y(:,1));
end
xTickAmount = sort(unique(x(1,:)));
if length(xTickAmount)<2
set(cax,'xtick',[]);
elseif length(xTickAmount)<=16
set(cax,'xtick',xTickAmount);
end %otherwise, will use xtickmode auto, which is fine
hold(cax,'off'), view(cax,3), grid(cax,'on')
set(cax,...
'NextPlot',next,...
'ydir','reverse');
if plottype==0,
set(cax,'xlim',[1-barwidth/m/2 max(x)+barwidth/m/2])
else
set(cax,'xlim',[1-barwidth/2 max(x)+barwidth/2])
end
dx = diff(get(cax,'xlim'));
dy = size(y,1)+1;
if plottype==2,
set(cax,'PlotBoxAspectRatio',[dx dy (sqrt(5)-1)/2*dy])
else
set(cax,'PlotBoxAspectRatio',[dx dy (sqrt(5)-1)/2*dy])
end
end
if nargout>0,
hh = h;
end
if(~colgroupcolors)
colorbar;
else
% Either the colorbar width is fixed or it doesn't automatically resize
% properly. I guess I'm stuck with the default width.
% caxpos = get(cax,'Position');
% caxpos(3) = caxpos(3)- 0.0;
% set(cax,'Position',caxpos);
UsedCmapMapInds= round([(1:nrows)' ,ones(nrows,1)]*cmapScalingT);
cmaphsv = rgb2hsv(cmap(UsedCmapMapInds,:));
cmaphsv(:,2) = 0.5 .^ SaturationPower;
cmaphsv(:,3) = SatToDecIntensity(cmaphsv(:,2));
showcmap = hsv2rgb(cmaphsv);
colormap(showcmap)
colorbarh=colorbar( 'YTick',[]);
a=reshape(showcmap,[size(showcmap,1),1,size(showcmap,2)]);
aI=repmat(a,[1,20,1]);
aIhsv=rgb2hsv(aI);
Smask = repmat(linspace(0,1,size(aIhsv,2)) .^ SaturationPower,size(aIhsv,1),1);
aIhsv(:,:,2) = Smask;
aIhsv(:,:,3) = SatToDecIntensity(aIhsv(:,:,2));
ScalingDataColorImg = hsv2rgb(aIhsv);
imageh=findobj(colorbarh,'Type','image');
XData = [1,nrows];
set(colorbarh, 'XLim',XData + [-0.5, +0.5]);
set(imageh, 'XData',XData);
YData = ScaleColorDataClim;
% For some odd reason I can't seem to set the limits and Xdata right.
% I figured it should be like an image which has XData from 1:n and
% Xlim from 0.5:n+0.5 but it doesn't seem to work. In fact for YLim it
% seemed to cause the tick labels on the histogram itself to go beserk.
% So, after sacrificing a chicken and consulting my oujie bard this is
% what I came up with.
% Is computer science an oxymoron or what?!
set(colorbarh, 'YLim',YData + [0.5, -0.5]);
set(imageh, 'YData',YData);
set(colorbarh,'YTickMode','auto')
set(imageh,'CData', permute(ScalingDataColorImg,[2,1,3]));
if(0) % an image of what a colormap for this should look like. I wish I could make a colormap out of this.
figure; image(ScaleColorDataClim,1:size(ScalingDataColorImg,1),ScalingDataColorImg);
end
end
function Int=SatToDecIntensity(Sat)
d=size(Sat);
Sat = Sat(:);
if(0)
si=...
[0,1;...
0.125 ./ 2, 0.99999;...
0.125, 0.9999;...
0.25, 0.996;...
0.5, 0.96;...
0.75,0.92;...
1, 0.75];
StoImap = [si(:,1) .^ 3, si(:,1) .^ 2,si(:,1), ones(size(si,1),1)];
StoImapT = StoImap \ si(:,2);
Int = max(0, min(1,[Sat .^ 3, Sat .^ 2, Sat, ones(length(Sat),1)] * StoImapT));
else
SquashPower = 0.4;
MinInt = 0.75;
Int = MinInt + (1 - MinInt) .* (cos(Sat .* 0.5 .* pi) .^ SquashPower);
end
Int = reshape(Int,d);
function params = mergedefaultparams(params, defaultparams)
if(isempty(params))
params = defaultparams;
return;
end
if(isempty(defaultparams))
return;
end
names = fieldnames(defaultparams);
for iname=1:length(names)
namei = names{iname};
if(~isfield(params,namei)) % add the default
params=setfield(params,namei,getfield(defaultparams,namei));
else % params has the field but if its a struct then the issue becomes recursive.
if(isstruct(getfield(defaultparams,namei)) & length(getfield(defaultparams,namei)) == 1 & ~isempty(getfield(params,namei))) % field name is in both so now may have to recurse.
params=setfield(params,namei, mergedefaultparams(getfield(params,namei),getfield(defaultparams,namei)));
end
end
end
