specifying the ydata for imagesc

I am trying to create an image of a matrix M in which the rows represent specific frequency values, the colums are time points and the values of the matrix (the cdata of the image) are the power at each time and frequency (in other words, it is a spectrogram). I have a vector Y of frequencies which describe the rows, and a vector X describing the times.

I use: imagesc(X,Y,M)

For good reasons, the values in my Y are not evenly distributed between Y(1) and Y(end). Matlab on the other hand ingores everything except Y(1) and Y(end) and assumes that the Y axis is evenly distributed bween the two. As a result, using e.g. 'ginput', or the data cursor tool to sample the image does not return the correct value of the Y axis (i.e., the correct frequency of that row in the matrix).

This behavior of ignoring everything but the start and end of Y is documented in the 'imagesc' help document, so it is a feature, not a bug, but it is annoying. I wonder if I a missing a way to get beyond this imiitation.

Thanks,

Leon

[Adi Natan]

you may want to try xyimagesc.m

(see the file exchange: http://www.mathworks.com/matlabcentral/fileexchange/32732-xyimagesc-m )

Answer 1

[bym]

perhaps:

imagesc(X,log(Y),M) %or whatever your mapping is

[Leon]

Thanks. The problem is that Matlab ignores whatever you put for Y, except its first and last element. So for instance Y = [2,4,8,16] and Y = [2,7,11,16] yield the same thing, both not reflecting the Y argument in the command. Try the following for example. You will get two similar images, and using the data cursor doesn't return the intended values in either one.

Leon

A = rand(4,10);

Y1 = [2,4,8,16];

Y2 = [2,7,11,16];

X = 1:10;

figure

subplot(2,1,1)

imagesc(X, Y1,A)

subplot(2,1,2)

imagesc(X, Y2,A)

Answer 2

[Walter Roberson]

For the data cursor, use a custom data cursor formatting function and map the linear y to the nearest desired Y (or an interpolation of it).

Answer 3

[Martin Kahn]

Seeing that this question still gets views despite its age:

One way of doing this is using surface(X,Y,Z) instead of imagesc. Don't forget to set the shading (i.e. shading flat;). To me it seems the Matlab documentation on this should be much clearer that imagesc simply can't do this.