以Micro-Structure Descriptor與Gray-Level Co-occurrence Matrix暨Particle Swarm Optimization為基礎之雙階層影像處理
陳宗信 · 2013
The media information on the internet increase quickly. Image retrieval has become an important topic. Keywords-based image retrieval (KBIR) consumes a lot of time and energy for tagging every image with keywords. Content-based image retrieval (CBIR), which makes it possible to retrieve images without keywords, has been proposed to retrieval image with image contents. It can avoid the consumption of time and energy. An original descriptor for CBIR called Micro-structure descriptor (MSD) mainly remains color features. The discrimination power of MSD feature is limit because MSD does not denote texture feature well. This thesis reduces MSD features and combines reduced MSD features with gray-level Co-occurrence matrix (GLCM). The GLCM denotes global features and texture features. Then, feature weights are used to fuse two categories of features so that the distance between two images is defined through weights. Since the distance between two images has been defined, the images in the same class can be divided into subclasses by K-mean clustering. Unlike traditional image retrieval methods, the proposed method will respond images in the same subclass for a query image. That is reason why the images in the same class should be further divided into subclasses. Finally, Particle Swarm Optimization (PSO) is used to optimize parameters, such as features weights and the number of subclasses in the same class. According to the experimental results, the proposed method can obtain better results for image retrieval.