Orientation angles for Haralick texture measurements

Haralick’s texture measurements (IEEE Transactions vol. SMC-3, no. 6, pp 610-621, 1973) can vary depending on the angle between the reference pixel and its neighboring pixel. That is, the value of a particular texture measurement can change depending on whether you take the neighboring pixel to be the pixel positioned to the right (at the 0 degree angle), or at a diagonal (45 or 135 degrees), or above (at 90 degrees) relative to the reference pixel.

Haralick dealt with this problem by calculating a new measurement for each orientation (0, 45, 90, and 135 degree angles), and so, for example, calculated four different values of the ‘angular second moment’ texture measurement for a single object. In contrast, CellProfiler calculates only a single value for any particular texture measurement. How does CP do this? Does CP assume that the orientation between reference and neighbor pixels to be 0 degrees, for instance? Or does CP average the four separate values for the different orientations?

Hi,

We have chosen to calculate the co-oocurence matrix using only an offset to the right (e.g., 0 degrees). We’ve had at least one other request to have CP either average across several angles or output a separate feature for each one. When we implement this feature, we’ll post it here.

Regards,
-Mark

Hi,

We’ve just included user-selection of the desired direction in MeasureTexture. You can wait until our next release to access these features, or if you’re daring, you can try it out with our latest public build from source code here. The trunk build is not as heavily vetted as our releases, so you should take that into account; please note the caveats mentioned on the page linked.

Regards,
-Mark