The Study Of Visual Sampling Clustering Method And Its Application In Image Retrieval

With the rapidly development of data mining and data analysis technologies, clustering analysis methods have been generally utilized in various fields such as pattern recognition, image processing, computing vision, and so on. The existed various clustering methods have been applied in different applications, with their distinctive advantages respectively. Nowadays, how to reduce the initial sensitivity while obtain the optimal cluster number, and properly explain the clustering procedure are some top issues that many scholars are addressing.This paper focuses mainly on some hot topics of the researches of clustering algorithms and its application. Our work includes the following subjects:1) Based on the visual sampling principle, the generalized visual sampling based clustering approach VSC is proposed. The clustering approach incorporates the visual sampling principle with the famous Weber law, such that it has two distinctive advantages: firstly, it is insensitive to initial conditions; secondly, the reasonable clustering number can be effectively determined by the new Weber-law-based clustering validity index. The experimental results demonstrate its success.2) After analysis the kernel algorithm SCA(Similarity-based Clustering Algorithm) which recently proposed by Yang Miin-Shen et al,. The link relationship between our approach VSC and algorithm SCA is derived. Both theoretic analyses and experimental results show that : in many cases, the approach here has almost the same clustering results as algorithm SCA. This fact reveals that the approach can be used to overcome the drawback of SCA, i.e., the parameterγis very difficult to be well determined.3) In content-based image retrieval, color features are widely used as important visual information in images. Compared with the RGB color space, the HSI color space is more acceptable in human visual attribute. Due to the flexibility and uncertainly of the image information, the single feature color in the HSI color space can be fuzzed into color histogram vector under the fuzzified mechanism. Based on the robust visual sampling clustering method, the fuzzy color histogram vector can be auto-clustered and the image can therefore be well matched in the color space.