An Analysis of Visual Interest Detection in 2D Game Concept Art

T. San ; Y. Kanematsu ; K. Mikami

Journal for geometry and graphics, Tome 27 (2023) no. 1, pp. 69-79.

Résumé

This study proposes a scheme that demonstrates the pattern of color combinations in the visual and non-visual centers. To achieve this goal, we first have to clearly distinguish the visual and non-visual center areas in the game concept design. Up to now, the saliency studies have been similar to our study. However, after our tests, the saliency studies do not distinguish well between the visual and non-visual center areas in the concept art. Therefore we propose this research method from the artist�s perspective. We first partition the image into small local regions, similar to bags of feature model, and then we give each local region a hash string to determine their similarity. Then we classify similar local regions. In the painting principle, similar or repeated regions are often non-visual centers. Finally, we check the variation of HSV of local regions to confirm the accuracy of the visual center. At the end of the study, we compared the performance of our study with two other major saliency methods in a sample of game concept designs. The results show that our method can better show the visual center of the game concept design diagram.

Détail

Classification : 91E10
Mots-clés : Visual interest, game concept art, color contrast, segmentation, 2D digital artwork