Discrimination of ceramic types using image processing

Abstract

Numerous areas of archaeological relevance have been excavated all over the Maltese Islands. During excavation, several different artifacts are recovered, which include pottery fragments. Some pottery fragments may have historical significance, hence it is of importance that these are analyzed. Analysis of pottery shards could give the archaeologists information about its age, origin of the materials used in the making of the pottery, as well as the function of the pottery. Hence, the pottery is analyzed under a microscope, from which inclusions need to be outlined. It is also of interest to the archaeologist that the pottery shard is not destroyed for analysis to be conducted. Hence, the pottery shard shall retain its shape as found in the historical site for the analysis done in this dissertation. This project also aims at automating the process of finding the inclusions which are of interest to the archaeologist.

The system will first enhance the image through histogram equalization to compensate for the effect of shadows due to the uneven surface of the pottery shard. Then the features for all the image pixels will be obtained. The features will be three colour spaces, namely the RGB space, the HSV space and the L*a*b* space. These features will be used to classify the pixels into different classes such that the inclusions of different colours could be distinguished from each other and from the matrix. The archaeologist will then be asked to choose the inclusion which is of interest, and that class will be isolated and noise will be removed. Finally, the information about the inclusions, such as size, compactness and eccentricity, will be obtained such that the archaeologist could perform further analysis.