Research: Body patterns & camouflage
In two-dimensional analysis, such as with images, the combination of low and high pass filters applied to the rows and columns of the image
matrix identifies directionaility in the texture patterns: low-pass to rows and high-pass to columns = horizontal; low-pass to columns and high-pass to
rows = vertical; high-pass to rows and columns = diagonal.
A small (200 x 200 pixel) patch of the dorsal markings of a Jacky lizard was analysed using a two level DWT (with the haar wavelet). The results are presented
on the right. H = horizontal coefficients, V = vertical coefficients, D = diagonal coefficients at the two levels. Not labelled is the approximation coefficient A.
(Note: it is possible to reconstruct the original image from the transform coefficients A2, D2, D1 using inverse DWT).
In order to compare coefficients, we can calculate the amount of 'energy' in each of the subbands (HVD), construct a feature vector of energy levels and
compare using a distance metric such as the Euclidean distance (e.g., Kiltie, Fan & Laine (1995) Mathematical Biosicences 126:21-39).
This is possible because texture energy in the wavelet domain is distributed differently for different texture patterns.
Return to previous page here
Home |
Research |
Last Update:
December 12, 2006