deltaEtAl

Calculates non-parametric contrast measures for any gray value image as described in the cited publication.

The function uses a single grey value image (matrix) as in input. In addition, a data structure thr is required as an input which contains threshold values obtained by Otsu's method. This structure is output by the attached function greyThreshO, which again calls the function multithresh from Matlab's image processing toolbox. (The users may furnish their own function that yields a similar data structure with thresholds as an output, in that case the image processing toolbox will not be needed). The optional input perc is only needed when the normalised slope at the thresholds (NST) is desired as an optional output, this will be described in a follow-up publication. Finally, the optional input pcov is needed if extreme values in the image (like defunct pixels which produce outliers) have to be excluded.

The function calculates the order statistic of the image and from that, with the help of the threshold(s), values of the relative grey value distance δ, the range R, the mean values LH and LL beneath and above the threshold as well as the pixel number nT where the threshold yT is located. If the input perc is nonzero, the normalised slope NST at the threshold yT is also calculated. To that end, the order statistic is scaled to the intervals [0,1] both for the grey values and the pixel number.

Using δ, one obtains the so-called mean brightness distance MBD by multiplication with R, i.e. MBD=. MBD divided by the sum of LL and LH yields the contrast equivalent distance CED = MBD / (LH+LL). For low noise, CED converges to the Michelson contrast .

For the calculation of the distance-to-noise ratio DNR, the additional function avThreshNoise is required, which is also available on the Matlab Central.(Please look for "avThreshNoise" or "Non-parametric Noise measure tau" ).

If the input thr.values contains more than one element, the contrast parameters are calculated for each threshold value separately. This is useful if more than two distinct grey value levels can be identified in the image.

For more details, please see the cited publication (free access).