The Research And Application On Fuzzy Clustering In Image And Video Fields

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, video data analysis and so on. The existing various clustering methods have been applied in different applications, with their distinctive advantages respectively. Nowadays, how to reduce the threshold sensitivity while obtain the optimal cluster number and how to reduce the influence of artificial operation are some hot issues that many scholars are addressing.This paper focuses mainly on some aspects of the researches of clustering algorithms and their application. Our work includes 3 pieces:1) In content-based retrieval, color features are widely used as important visual information in images. We will take the SCD and CLD these two colour descriptors as the visual features for the shot description. Due to the flexibility and uncertainly of the image information, the single feature color in the HSY color space can be fuzzified into color histogram vector under the fuzzified mechanism, and we can use the histogram in the similarity measures. With these colour descriptors, we gain the good effect.2) After research on-line clustering of unsupervised clustering, a generalized fuzzy clustering approach MRLC is proposed. This clustering approach incorporates the unsupervised clustering algorithm Leader-follower with the concept of radius and center, such that it has two distinctive advantages: firstly, it is insensitive to initial conditions; secondly, it is suitable for large data analysis, e.g. the video data. The experimental results demonstrate its success.3) After analysis the kernel algorithm Leader-follower which is based on the similarity measure, we use other fuzzy clustering approaches to modified MRLC more. When doing the key frame extraction and key shot extraction, in term of practice requirements, the new approaches get improved separately. It is more acceptable in human visual attribute. At the same time, the thresholds are used with fewer times and with less manual operation.