Multi-contrast based texture model for understanding human subjectivity

We study the subjective interpretation for statistical textured images in order to clarify the human subjective interpretation mechanism for images. We focus on the orientation interaction and color property, and unified those features by calculating the contrast over the entire image resolution. We measure the psychological responses and correspond them to our texture feature based on the canonical correlation statistics. We develop an image retrieval system for textured images based on our model. The results obtained show that our model can retrieve images which give a similar impression, and can also predict images from some descriptive adjectives