Showing a new image inside existing axes w/o deleting the colorbar

Question

I am working on a GUI that is being initialized by creating several axes along with an invisible colorbar for each one (this is done so that axes maintain their predefined Position)¹. The handles to all axes and colorbars are stored.

Interactions with the UI may result in images being plotted on any one axes. I would like only the colorbar of the active axes to be shown at any given time, by appropriately setting the Visible property of all colorbars.

I am having a problem with this approach, because I'm using imagesc to update my axes, and this deletes any colorbar associated with the axes, making the stored handles invalid.

My question is: how can I use imagesc or image to update an axes that is associated with a colorbar, without the colorbar being deleted?

Here's how to reproduce this problem:

dbclear in newplot %// Needed for the code to be properly re-runnable
%// Create an example figure containing a colorbar:
figure(); imagesc(imread('cameraman.tif')); colorbar;
%// "Refreshing" the displayed image:
uiwait(msgbox('The image will now be refreshed. A breakpoint will be set in newplot.m'));
dbstop in newplot at 124 %// The line responsible for deleting the colorbar in R2015A/B
imagesc(imread('cameraman.tif'));

The line on which the breakpoint in newplot.m is set reads:

cla(ax, 'reset', hsave);

Which is (un)surprisingly an undocumented way to invoke cla (with 3 parameters) that preserves objects whose handles are found in hsave.

---

Some ideas I had assuming colorbar deletion is inevitable (which I will pursue in case a "sane" solution is not found):

1. Binding a DeleteFcn to the colorbar that saves its data to some struct. Creating a new colorbar after imagesc has finished, then iterating over the fields of the struct and assigning all properties to the new colorbar object.
2. Check "every once in a while" that all colorbars exist using either findall(hFig,'type','colorbar') or verify that each of the axes has a valid ColorbarPeerHandle as per the Appendix below. Recreate the CB if invalid.
3. Delete all colorbars whenver different axes become active and create only the CB I'd like to show.

---

Appendix - ColorBar/Axes association:

• Handle to the colorbar associated with a certain axes hAx can be obtained (in hg2) using:

   hCb = getappdata(hAx,'ColorbarPeerHandle');
  

• Handle to the axes associated with a colorbar object hCb can be obtained using²:

   hAx = hCb.Axes;
  

Answer 1

I've managed to come up with a couple of solutions:

1. safeUpdateImage1 - Based on the circumvention of cla(...) via the creation of a new Image object inside the axes after using hold. This is good for cases when an Image doesn't necessarily exist in the axes.
2. safeUpdateImage2 - Following mikkola's suggestion, based on updating an existing Image object. This is good for cases when there already is a ColorBar and an Image associated with the axes.
   

Here's a demonstration of both solutions alongside the original problem:

function [] = Problem
dbclear in newplot %// Needed for the code to be properly re-runnable
%// Create an example figure containing a colorbar:
Img = imread('cameraman.tif');
figure(); imagesc(Img); hAx = gca; colorbar;
%// Refreshing the displayed image (comment/uncomment as needed):
switch questdlg('Please select an image update method:','Update method selection',...
                'Broken','Safe1','Safe2','Safe1')
  case 'Broken'
    brokenUpdateImage(hAx,255-Img);
  case 'Safe1'
    safeUpdateImage1(hAx,255-Img);
  case 'Safe2'
  safeUpdateImage2(hAx,255-Img);  
end
end

function brokenUpdateImage(hAx,newImg)
uiwait(msgbox('The image will now be refreshed. A breakpoint will be set in NEWPLOT'));
dbstop in newplot at 124 %// The line responsible for deleting the colorbar in R2015A/B
imagesc(newImg,'Parent',hAx);
end

% For cases when the only desired child is an Image and the axes contents are unknown
function safeUpdateImage1(hAx,newImg,imgType,imgUpdateFcn)
if nargin < 4 || isempty(imgUpdateFcn)
  imgUpdateFcn = @imagesc;
end
if nargin < 3 || isempty(imgType)
  imgType = 'Image';  
end
if strcmp(hAx.NextPlot,'replace') %// Equivalent to checking "ishold == false"
  hAx.NextPlot = 'add'; %// Equivalent to "hold on"
  %// hCurrImgs = get(hAx,'Children'); %// Deletes all types of Children
  hCurrImgs = findall(hAx,'type','Image'); %// Deletes only graphical objects of type 
                                           %// "matlab.graphics.primitive.Image"
  for ind1=1:numel(hCurrImgs)
    delete(hCurrImgs(ind1));
  end
  imgUpdateFcn(newImg,'Parent',hAx);
  if strcmpi(imgType,'Image')
    axis(hAx,'tight','ij');    
%// 'tight' - XLimMode, YLimMode, and ZLimMode change to 'auto'. The limits automatic- 
%//           ally update to incorporate new data added to the axes. To keep the limits
%//           from changing when using hold on, use axis tight manual.
%// 'ij'  —   Reverse direction. For axes with a 2-D view, the y-axis is vertical with
%//           values increasing from top to bottom.    
  end
end
end

%// When it's known that the axes contains at least one Image:
function safeUpdateImage2(hAx,newImg,~,~)
%// <Input checking code>
hCurrImgs = findall(hAx,'type','Image');
%// <Optional code to decide which Image child to update>
hCurrImgs(1).CData = newImg; %// In this example, only update the "topmost" child
end