Art Pricing with Computer Graphic Techniques

TL;DR

This paper introduces computer graphics techniques to art pricing, using image recognition to quantify painting effort and improve valuation accuracy in a novel way.

Contribution

It pioneers the use of image recognition for measuring artist effort, enhancing traditional hedonic models in art pricing research.

Findings

Variances of lines and colors significantly influence sales prices.

Proposed measurements better capture content heterogeneity.

Method offers a new quantitative approach for valuation and authentication.

Abstract

This paper makes the first attempt to introduce the tools from computer graphics into the art pricing research. We argue that the creation of a painting calls for a combination of conceptual effort and painting effort from the artist. However, as the important price determinants, both efforts are long missing in the traditional hedonic model because they are hard to measure. This paper draws on the digital pictures of auctioned paintings from various renowned artists, and applies the image recognition techniques to measure the variances of lines and colors of these paintings. We then use them as the proxies for the artist's painting effort, and include them in the hedonic regression to test their significance. Our results show that the variances of lines and colors of a painting can significantly positively explain the sales price in a general context. Our suggested measurements can better capture the content heterogeneity of paintings hence improving on the traditional art pricing methodology. Our approach also provides a quantitative perspective for both valuation and authentication of paintings.