You can specify x/y for imagesc so that the image is not shown against its index. From there you can simply use hold on and contour:
y = [1:10]';
x = 1:0.01:10;
z = repmat(y, [1,901]) - repmat(x, [10,1]);
figure
subplot(2,1,1)
contourf(x,y,z)
xlabel('x (continuous)');
ylabel('y (discrete)')
subplot(2,1,2)
imagesc(x,y,flipud(z)) %specify x/y
xlabel('x (continuous)');
ylabel('y (discrete)')
colormap(jet(256))
hold on %hold it on
contour(x,y,flipud(z),'color','k') %contour over it
More per Comments
In that case I would grab the 'CData' from the image and bin it using histc (essentially replacing image values with indexes into contour levels). Then draw a contour plot of this.
y = [1:10]';
x = 1:0.01:10;
z = repmat(y, [1,901]) - repmat(x, [10,1]);
figure
subplot(2,1,1)
contourf(x,y,z)
xlabel('x (continuous)');
ylabel('y (discrete)')
subplot(2,1,2)
h = imagesc(x,y,flipud(z));
xlabel('x (continuous)');
ylabel('y (discrete)')
colormap(jet(256))
hold on
D = get(h,'CData'); %image data
[~,idx] = histc(D,linspace(m(min(D)),max(max(D)),10)); %bin it
contour(x,y,idx,'color','k') %contour of the bins
