This study explores the use of a generic shape descriptor for quantitative comparisons between the full-field strain data acquired from virgin and damaged composite panels using digital image correlation. These descriptors are capable of decomposing images with 10 3 to 10 6 pixels into a feature vector with less than a few hundred elements. Strain distributions in four composite specimens with incremental impact damage and a virgin specimen, all subject to a tensile load, were decomposed using the newly developed Fourier-Zernike descriptors. Pearson's correlation coefficient, cosine similarity and Euclidean distance were employed to compare quantitatively the feature vectors evaluated for the strain distributions in the four damaged specimens with the strain distribution in the virgin specimen. The deviation of the Pearson correlation coefficient from unity was found to be an effective damage indicator, which could be evaluated automatically and without the need for subjective assessment of the damage.