Tomoharu SATO¹, Takehiro NAGAI^2*, and Shigeki NAKAUCHI¹

¹Department of Computer Science and Engineering, Toyohashi University of Technology, Toyohashi, Aichi 441-8580, Japan
²Department of Informatics, Yamagata University, Yonezawa, Yamagata 992-8510, Japan

(Received February 12, 2014; Revised April 11, 2014; Accepted July 3, 2014)

We applied the classification image (CI) method to examine the effects of heterochromatic noise on color perception. Moreover, rather than the typical CI analysis procedure, we analyzed the CI data based on chromatic mechanism models. The stimulus was a superposition image of two uniformly colored squares (signal images) and a multicolored texture (a noise image), whose colors were chosen from an isoluminant plane of the Derrington–Krauskopf–Lennie color space. The observers judged the relative chromatic contrasts of the two signal squares on the different noise textures. The CI showed strong color modulations, whose color directions differed from the signal. Additionally, the model analysis demonstrated that the model with four mechanisms and the cardinal mechanisms model were not inferior to the model with more mechanisms with regard to explaining our experiment data; the cardinal model's fit with the observer's behaviors was improved by simply adjusting the relative sensitivities of the four mechanisms.

Key words: color vision, higher order chromatic mechanisms, classification images

^*E-mail address: tnagai@yz.yamagata-u.ac.jp