A computational approach for analysis of art compositions

Abstract

New approaches emerging from the analysis of artworks with computational tools have the potential to offer different perspectives to artworks recreated in digital environments. This study aims to reveal the implicit relationships between Mondrian compositions with different visual representations. In the scope of the study, compositions completed between 1938 and 1943, which have a strong geometry-color relationship, were first investigated through a pixel-based approach. In the fragmentation method followed, the similarities and differences are expressed with data transferred from pixels to numerical matrices in two different steps: 1. Between the artifacts in pairs, 2. Between an artifact and all the other selected artifacts. The visualization of the matrices was represented by 2D color maps and 3D texture maps. These interpretation styles allow the compositions to be expressed from general to specific and again, from specific to general, by gaining a new meaning.