DoG and salient features-based scheme for improving quality of visual words

作者：Li Weisheng^*; Liu Rui; Huang Ying; Zhou Lifang

来源：International Journal of Advancements in Computing Technology, 2012, 4(15): 379-386.

DOI：10.4156/ijact.vol4.issue 15.44

• Bag-of-visual words
• Clustering
• Computer vision problems
• Extraction method
• Keypoint
• Keypoints
• Recognition rates
• Semantic information
• Spatial informations
• Visual word
• Word models
• X-means

With the rapid development of bag-of-visual word model and its wide-spread applications in various computer vision problems, it is an increasingly important issue to improve the quality of visual words generated from extracted keypoints. In order to do this, approaches that combine different keypoint extraction methods are proposed by many researchers. However, there are two major drawbacks for the available combination: some noisy keypoints are extracted and the semantic information between the keypoints is ignored. This paper presents a framework to overcome these limitations. First, the difference-of-gaussian (DoG) and Salient detector are combined to obtain a large set of possible interest points in an image. Then the X-means clustering algorithm which utilizes the spatial information of keypoints is used to reduce the size of the keypoints set. Finally, a comprehensive comparison is made between the proposed method and the classic method. The experimental results show that the novel method can effectively optimize the keypoints and improve the recognition rate of objects.

• 出版日期2012