Aesthetic Quality Classification of Photographs Based on Color Harmony

We treat a photograph as a collection of local regions with relatively simple color variations and use color harmony as a feature. This enables aesthetic quality classification even for photographs with complex color arrangements.

This study explores how a computer can identify photographs that people find beautiful in a large collection. One factor is color harmony: how well the colors work together. Existing models mainly consider simple combinations of a few colors, making it difficult to evaluate a photograph with a complex color arrangement.

We address this by treating a photograph as a collection of small regions. Even when the whole image is complex, color changes within each region are relatively simple. We summarize the color patterns found in these regions and use them as clues to the photograph's aesthetic quality. We call the method bags-of-color-patterns.

In experiments on a large collection of photographs rated by people, the method performed better than the existing methods used for comparison. Classification improved further when we combined color harmony with information about blur, edges, and visually noticeable areas. The results show the value of considering several aspects of a photograph alongside its colors.

• Masashi Nishiyama, Takahiro Okabe, Imari Sato, and Yoichi Sato,
  Aesthetic Quality Classification of Photographs Based on Color Harmony,
  Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), June 2011.

• 西山 正志, 岡部 孝弘, 佐藤 いまり, 佐藤 洋一,
  審美的品質識別のための局所領域の組合せによる色彩調和の評価,
  電子情報通信学会論文誌 D, Vol.J94-D No.8 pp.1324-1334, 2011.