Reliability of textural analysis of ancient plasters and mortars through automated image analysis

Abstract-For the study of ancient or recent plasters and mortars, it is necessary to determinate some morphological and textural parameters to describe the aggregate fraction and its relationship with binder fraction. Textural analyses are commonly performed with simple visual comparators or, at most, with a time-consuming manual point counting. The introduction of automatic or semiautomatic image analysis techniques seems to have made this kind of determinations faster and less biased. To compare data coming from different image-processing sequences or different analytical approaches, reliability and calibration need to be tested by each operator. A critical principle setup is presented for comparison using laboratory mortars with well-known textural features; classical analysis by sieving, point-counting technique and automated image analysis were performed. This empirical strategy allowed (1) to develop a simple and easy-to-use software-based image analysis system for plaster and mortar samples; (2) to pull out the relations between samples textural features and reliability of the different approaches; (3) to provide linear relations that allow to convert apparent measured features to real features. Results also underline the need to calibrate and standardize image analysis techniques before data handling.