MATLAB overflows (i.e. produces an "Out of Memory" error) when attempting to render a surface from an array with a large number of data points for the following reason:
When creating a surface using the SURF function, MATLAB creates a copy of the data arrays. If there is not enough contiguous memory available to store these arrays, it will result in an overflow. For more information regarding MATLAB memory allocation, refer to the Memory Managament Guide technical note (tech note 1106) on The Mathworks Support web page.
If there is enough contiguous memory available, then the overflow is a result of the amount of memory required to render the surface.
To represent the this surface with a large number of data points requires (9851-1)*(6916-1) = 68 million quadrilaterals. Quadrilaterals are rendered by graphics engines as two triangles. Thus this particular surface requires 136 million triangles to be rendered. Each triangle is represented by 3 floating point numbers per vertex for the X, Y, Z coordinates. This works out to 4.9 gigabytes (GB) of data not including other information about colors, edges, and normals. This amount exceeds the memory available to MATLAB 7.1 (R14SP3) on a Windows platform.
To work around this issue, try reducing the amount of plotted data to produce a surface plot representative of the data.
Here is one way to downsample and plot the data:
[u v] = size(A);
t1 = 1:10:u;
t2 = 1:10:v;
b = A(t1,t2);
surf(b)
