A content-based image retrieval system for hat database

Content-based image retrieval (CBIR) is a technique for indexing and retrieving images based on the low-level features, middle-level features, and high-level features. Low-level feature is extracted from contents of the images such as color, texture and shape; middle-level feature is a region obtained as a result of image segmentation; high-level feature is semantic information about the meaning of image, its objects and their roles, and categories to which the image belongs. In this project, three low-level features texture-based retrieval, color-based retrieval and shape-based retrieval are implemented and compared on hat database. Texture features are obtained from parameters of a two-component Gaussian mixture model (GMM) in the wavelet domain. Color features are extracted from a two-component GMM on HLS color space. Shape features are extracted from the contour by using centroid-contour distance Fourier descriptor. A comprehensive experimental evaluation of the retrieval performance of different feature sets is performed. The experimental results indicate that the shape features based on the centroid-contour distance Fourier descriptor perform much better than the color and texture features for the hat database used in this project

A content-based image retrieval system for hat database

Content-based image retrieval (CBIR) is a technique for indexing and retrieving images based on the low-level features, middle-level features, and high-level features. Low-level feature is extracted from contents of the images such as color, texture and shape; middle-level feature is a region obtained as a result of image segmentation; high-level feature is semantic information about the meaning of image, its objects and their roles, and categories to which the image belongs. In this project, three low-level features texture-based retrieval, color-based retrieval and shape-based retrieval are implemented and compared on hat database. Texture features are obtained from parameters of a two-component Gaussian mixture model (GMM) in the wavelet domain. Color features are extracted from a two-component GMM on HLS color space. Shape features are extracted from the contour by using centroid-contour distance Fourier descriptor. A comprehensive experimental evaluation of the retrieval performance of different feature sets is performed. The experimental results indicate that the shape features based on the centroid-contour distance Fourier descriptor perform much better than the color and texture features for the hat database used in this project

Application of Shape Context and Fourier Descriptor to Shape Description in Preparation of Nanowood Powder

The nanowood powder made from varying wood species has different properties. Therefore, while preparing nanowood powder, it is critical for us to accurately identify the wood species. By adopting the method of analyzing the wood species from the microstructure of the plate cell, we are able to ensure a high level of accuracy of the recognition. It is one of the commonly used parameters to identify sheet materials for further exploration of the microscopic cell shape. While applying the method of improved Fourier descriptor and the method of shape context, we have carried out the mathematical analysis of common soft wood plank cells in 5 slice images. In addition, we have conducted studies on the single cell graphics and quadrilateral, pentagon, round, oval similarity analysis respectively, and explored the use of benchmark cell model, thus providing statistical support for the subsequent material identification. Experiments have shown that this method is both efficient and robust in identifying sheet materials.

Combining quantitative analysis with an elliptic Fourier descriptor: A study of pottery from the Gansu-Zhanqi site based on 3D scanning and computer technology

Pottery is an important material in archaeological studies, and the accurate classification of pottery shapes largely depends on the experience and knowledge of archaeologists. In this thesis, pottery taken from the Gansu-Zhanqi site is used for sampling. Three-dimensional (3D) models of the pottery were obtained using 3D scanning, and a computer-assisted pottery typology was studied through quantitative analysis and elliptic Fourier descriptor. This method, which can enhance and supplement the traditional methods of classifying pottery in archaeology and thereby enrich the parameters and breadth of pottery analysis, represents a new means for exploring and experimenting with objective classification and provides a new tool for traditional archaeological analysis methods.