不同空間頻率的情景對於物體辨識的影響

Object Identification in Spatially Filtered Scene Background

朱慶凡(Ching-Fan Chu)

指導教授 ： 葉怡玉

國立臺灣大學/理學院/心理學研究所/碩士(2010年)

https://doi.org/10.6342/NTU.2010.03397

摘要

過去研究認為低頻情景活化出情景的結構有助於物體辨識（Bar, 2004）；而本研究的目的在於檢視這樣的假設是否成立，藉由操弄不同空間頻率的情景訊息，檢視情景訊息如何影響物體辨識。實驗一 A裡物體與情景圖片經由相同的濾器處理，實驗中物體呈現時間為36毫秒並且可能單獨出現或是伴隨著情景出現。結果發現物體單獨呈現時物體辨識表現優於物體呈現在情景之中，顯示情景訊息干擾物體辨識；實驗一 B裡刺激呈現時間延長至受試者反應仍得到相似的結果。情景干擾物體辨識可能是旁側抑制產生的結果，因此實驗二裡物體與情景圖片分別經由不同的濾器處理。結果發現刺激呈現36毫秒時當情景與物體圖片裡共享的空間頻譜越廣物體辨識表現越差，顯示情景產生的旁側抑制會干擾物體辨識。實驗三裡，比較物體呈現在雜訊背景或是情景背景中的辨識表現，檢視受到相同的旁側抑制影響時情景訊息是否可以幫助物體辨識。結果發現在低頻情景裡的物體辨識表現比在低頻雜訊裡的好，顯示低頻的情景有助於物體辨識。在本研究裡我們發現低頻的情景有助於物體辨識，但特定空間頻率（約3 cpd）的物體訊息仍是影響物體辨識表現的關鍵，透過整合特定空間頻率的物體訊息和低頻的情景訊息才能使物體辨識表現達到最好。

關鍵字

物體辨識 ； 空間頻率 ； 情景知覺 ； 旁側抑制

並列摘要

Identification of objects in scene contexts is critical for behavioral actions. Yet, how scenes influence object identification remains uncharacterized. It has been suggested that the low-spatial frequency of scene context is beneficial for object identification through activating possible objects in the context frame subserved by parahippocampus (Bar, 2004). In the present study, we verified this model by investigating the influence of spatially filtered scenes on object identification. In three experiments, objects were the ones that frequently occurred in the scenes and hence, were semantically and thematically congruent. In Experiment 1A, objects and scenes were spatially filtered by the same cutoff frequency (low- or high-passed at 1, 3, or 5 cpd). Objects were presented for 36 ms either alone without scene background or in scene contexts. Results showed that scene decreased the performance of object identification. The same results were obtained in Experiment 1B where stimuli were presented until response. In Experiment 2, we investigated whether the decrement results from lateral masking by manipulating the degree scenes shared the same frequency components with the objects. Scenes were filtered (low- or high-passed) at six cutoff frequencies (0.4, 0.7, 1.4, 2.8, 4, or 5.6 cpd) and objects were filtered (low- or high-passed) at 2 cpd. The results showed that when the object and the scene shared more spatial frequency band the identification performance became worse after viewing stimuli for 36 ms. These results supported that scene can produce the effect of lateral masking that reduces the performance of object identification. In Experiment 3, we used the same cutoff frequencies as in Experiment 2 to filter background and presented intact objects without filtering. Viewing duration was 36 ms. We examined whether scenes are beneficial by comparing performance with object identification in noise background where the frequency components were the same as in the scenes and lateral masking occurred in both types of background. The results showed that object identification in scene background was better than that in noise background. In addition, the scene benefit was eliminated when the low spatial frequency was removed from the scene. Taken together, the results supported that the low-spatial frequency of scene contexts is beneficial for object identification. However, objects must contain high spatial frequency components so that lateral masking within frequency channels does not lower identification performance. Activation of possible objects is beneficial only when details of objects can be integrated for identification.

並列關鍵字

spatial frequency ； object identification ； scene perception ； lateral masking

參考文獻

Adini, Y., Sagi, D., & Tsodyks, M. (1997). Excitatory-inhibitory network in the visual cortex: Psychophysical evidence. Proceedings of the National Academy of Sciences of the United States of America, 94, 10426-10431.

Bar, M. (2003). A cortical mechanism for triggering top-down facilitation in visual object recognition. Journal of Cognitive Neuroscience, 15, 600-609.

Bar, M. (2004). Visual objects in context. Nature Reviews Neuroscience, 5, 617-629.

Bar, M., Aminoff, E., & Schacter, D. L. (2008). Scenes unseen: The parahippocampal cortex intrinsically subserves contextual associations, not scenes or places per se. The Journal of Neuroscience, 28, 8539-8544.

Biederman, I. (1972). Perceiving real-world scenes. Science, 177, 77-80.