Identification of artistic styles using a local statistical metric

An algorithm for identifying the artist who created a picture is described. The algorithm relies upon computing the distribution of long and short lines in the image and comparing this distribution. The algorithm is one example of algorithms which can be designed to answer questions about global characteristics. This particular example computes the averages against a precomputed set from model samples. The distribution of line lengths is a statistical characterization which can be used to distinguish between artists. The current implementation is limited to black-and-white binary images. The results of testing the implementation on the daily comics is presented. Some of the related work in both computer science and art history which provides a conceptual background for the algorithm is also discussed.<
>