The conventional scene categorization methods ignore spatial information within a scene and are not able to discern categories that share similar subscenes but different in layout; or categories that are ambiguous by nature. To address this issue, in this paper a method is proposed to incorporate subscene attributes within global descriptions to improve categorization performance, especially in ambiguity cases. This is done by encoding subscenes with layout prototypes that capture the geometric essence of scenes more accurately and flexibly. The proposed method improves categorization accuracy. the proposed method can detect and evaluate ambiguity images more accurately.In this paper, scene categorization method is proposed by including subscene attributes to global descriptors. The use of prototypes is proposed to model the geometric configuration of subscenes. These prototypes are more accurate at capturing the layout and simple in training compared to the shape element-based approaches. Incorporating subscene descriptors can enhance the scene categorization result. Having been capable to detect ambiguity, the proposed method offers better understanding of the scene.
[1]
Krista A. Ehinger,et al.
Basic level scene understanding: categories, attributes and structures
,
2013,
Front. Psychol..
[2]
N. H. C. Yung,et al.
Improve scene categorization via sub-scene recognition
,
2014,
Machine Vision and Applications.
[3]
Alberto Del Bimbo,et al.
Local Pyramidal Descriptors for Image Recognition
,
2014,
IEEE Trans. Pattern Anal. Mach. Intell..
[4]
Ping Guo,et al.
Combining Local Binary Patterns for Scene Recognition
,
2014,
J. Softw..
[5]
Z. Jane Wang,et al.
An Unsupervised Hierarchical Feature Learning Framework for One-Shot Image Recognition
,
2013,
IEEE Transactions on Multimedia.
[6]
Joo-Hwee Lim,et al.
Extended Spectral Regression for efficient scene recognition
,
2014,
Pattern Recognit..