CBIR-MSVM: Content-based Image Retrieval using Multi-Labelled Support Vector Machines
Justiner Joseph, Xuewen Ding
Abstract: Abstract: Content Based Image Retrieval (CBIR) technique is the emergent application to extract the appropriate query based images. But, the query based extraction is the one of complicated task for reducing the classification accuracy. To overcome these issue, proposed the CBIR based Multi-labelled Support Vector Machine classifier is used to enhance the classification outcomes. The preprocessing stage is processed into two main forms such as image resizing and the image filtering. In this framework, Gaussian Filtering technique is performed to remove the unwanted features and filter the relevant content based features. Then, three feature extraction process are as color, shape, and the texture feature are extracted based on the Color Histogram, REGIONPROPS, and the Grey Level Co-occurrence Matrix (GLCM). The Color Histogram technique is utilized to remove the unwanted RGB based structures from the results of preprocessed image and applying REGIONPROPS shape feature to extract the specific area and the perimeter based shapes. Then, performing GLCM texture based feature extraction to extract the statistical related features. Among the extracted features, the similarity computation process is accomplished to classify the content based images. Finally, MSVM classifier is processed to classify the content based pictures. The presentation result of the proposed framework is predicted with the help of parameters such as precision, specificity, recall, sensitivity, and the classification accuracy. Hence, the proposed research work is superior to the other existing techniques.
Keywords: Content Based Image Retrieval, Multi-labelled, Histogram, Texture feature, Shape, Color., M.Tech / M.E / PhD Thesis, Engineering, China
PDF Link: https://www.ijsr.net/archive/v7i1/ART20179252.pdf