Purchase this article with an account.
Masamitsu Harasawa, Takao Sato; The critical factor for performance improvement in multi-frame orientation texture segregation. Journal of Vision 2003;3(9):616. doi: https://doi.org/10.1167/3.9.616.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Sequential presentation of multiple frames with the same texture characteristics improves texture segregation. In this study, we evaluated contributions of three potential factors for this improvement (border-signal enhancement, internal-noise reduction, and narrowed tuning for border detection) by using internal noise estimation method based on perceptual template model. Methods. Stimuli were dynamic texture consisting of 8 × 8 Gabor patterns. The whole field was divided into two areas of homogenous texture characteristics by either a horizontal or vertical border. Subjects' task was to identify the orientation of border (2AFC). Within each texture region divided by the border, orientations of Gabors were varied randomly following a normal distribution. The mean orientation for each region was randomized between frames while keeping the difference between regional means constant. Each texture frame was presented for 100ms without ISI. Number of presented texture frames was from 1 to 5. The thresholds for difference of mean orientations (signal) were measured for each value of standard deviation of intra-region orientation distribution (external noise) using a QUEST method. Results. The threshold decreased as external noise was decreased and as the number of presented frames was increased. Discussion. The results of analysis indicated that the performance improvement can be attributed to both enhancement of textural border signal and noise reduction by multiple sampling. The contribution of signal enhancement suggested that an intermediate representation of textural border is pooled over frames. The effect of internal noise reduction seems to be accounted for by the effect of probability summation by multiple sampling. These results indicate that border-signal enhancement and internal-noise reduction may contribute together for the improvement, but the effect of narrowed tuning is minimal.
This PDF is available to Subscribers Only