Emergent recognition of objects hidden in degraded images in the absence of explicit top-down information T Murata1, T Hamada2, T Shimokawa1, M Tanifuji3, T Yanagida4 |
|---|
1Center for Information and Neural Networks, National Institute of Information and Communications Technology, Japan |
| It is well known that recognition of severely degraded images such as two-tone ‘Mooney’ images is facilitated by top-down processing, in which priorly given information about the hidden objects play an effective role in recognizing the defective object images. Even in the absence of any explicit top-down information, however, we can still recognize the hidden objects during continued observation of the images in an emergent manner accompanied by a similar feeling to ‘Aha!’experience. Neural mechanisms of this kind of recognition without the top-down facilitation are poorly understood. Since this phenomenon is characterized by longer latencies ranging in seconds, we measured time for subjects to recognize objects hidden in degraded images. We found that the time follows a particular exponential function related to severity of image degradation and subject’s capability, which could be determined independently each other. This function was well accounted for by a theoretical model based on feature-combination coding of visual objects, in which neurons representing the object's features removed by the image degradation show stochastic activation to complement the representation of the object to be recognized. The present results suggest that the stochastic process working on feature combination coding of objects underlies the emergent recognition. |
| Up Home |
|---|