灰度
人工智能
面子(社会学概念)
计算机视觉
面部识别系统
计算机科学
感知
突出
低分辨率
模式识别(心理学)
心理学
高分辨率
图像(数学)
社会科学
遥感
神经科学
社会学
地质学
作者
Andrew W Yip,Pawan Sinha
出处
期刊:Journal of Vision
[Association for Research in Vision and Ophthalmology]
日期:2010-03-15
卷期号:2 (7): 596-596
被引量:34
摘要
One of the key challenges in face perception lies in determining the contribution of different cues to face identification. Here we focus on the role of color cues. Although color appears to be a salient attribute of faces, past research has suggested that it confers little recognition advantage for identifying people. A possible reason for the observed lack of a contribution of color to face recognition in these studies is that in situations where strong shape cues are available (as in high-resolution face images), performance may already be at ceiling and the contribution of color may not be evident. An interesting condition to examine, therefore, is one wherein shape cues are progressively degraded. The contribution of color to face recognition, if any, would be more evident under such conditions. To this end, we tested subjects' recognition performance with a set of famous faces that had been low-pass filtered at different thresholds. The faces were presented either in full-color or in grayscale. As in previous studies, we found no difference in performance for the color and grayscale conditions at high resolution; however, performance with color was significantly better than performance with grayscale images at lower resolutions. Our next experiment was designed to determine whether color contributes to face-recognition by providing diagnostic cues to identity or by facilitating low-level image analysis, such as segmentation. We tested subjects' performance with pseudo-color images that, we reasoned, would disrupt diagnostic cues but not low-level ones. We found performance with such images to be on par with true-color images. Taken together, our experimental results suggest that color does contribute to face recognition, and it does so by aiding low-level image analysis.
科研通智能强力驱动
Strongly Powered by AbleSci AI