I don't know if there's a generalized canonical way to compare the similarity of graphics objects, but they're going to differ.
% fake data
x = randn(1000,1);
% two objects of the same type, using the same data
h = histogram(x); hold on
g = histogram(x);
% they're not equal
isequal(h,g)
% find where props differ
% this includes properties which aren't public
% though i'm only checking one layer deep
hprops = struct2cell(struct(h));
gprops = struct2cell(struct(g));
equalprops = false(size(hprops));
for k = 1:numel(hprops)
equalprops(k) = isequal(hprops{k},gprops{k});
end
fnames = fieldnames(struct(h));
% a list of the fieldnames and values where they differ
d = [fnames(~equalprops) hprops(~equalprops) gprops(~equalprops)]
Things like handles to their hidden descendant objects will differ. The colors of objects (e.g. line objects in a plot) will differ, and their level in the ui stack will differ.
I'm assuming that a test for equivalence will have to consider which properties actually matter and compare those properties alone.
