| dc.description.abstract | The diffusion of computing techniques within humane studies is considerably increasing, due to technological advances and reduction in computing costs, with History of Arts being one of the most interesting fields of application. In fact, several domains related to works of arts might be investigated other than archival and retrieval of information: this includes the planning for the preservation and restoration, the developing of critical examinations of artworks as well as the supporting of less specialized activities such as teaching, presentation of museum exhibitions and spreading of information.Within this framework, it is of particular interest to review already known image compression techniques from other fields and to evaluate their applicability taking into account the peculiarities and the requirements of this specific domain. A distinguishing aspect of History of Arts is the requirement to keep the maximum approximation of the original information on each point of an image. This peculiarity strongly requires that the set of pixels of the same color in the original image is bound to a unique color in the final image and that each color in the final image uniquely determines the set of colors it represents. This requirement constrains the applicability in this domain to those quantization algorithms satisfying this property while still minimizing the quantization error.The paper gives a formal definition of the above requirement and justifies the use of the quantization and dithering algorithms based on the octree reduction technique. Following, the results given by the application of these techniques to several paintings are shown. Finally, the measured quantization errors are compared with those obtained by applying the median-cut algorithm. | en_US |
