Applying CBIR to interior decorating.

摘要

This project explores the use of content based image retrieval in the domain of interior decorating. The images in the database represent items used for interior decorating such as wallpapers, borders, fabrics, floorings, and carpets.; A comparison is done on the performance of different variations of the colour and texture features and of various distance metrics used to measure the similarity between the query and database features. The colour features extracted from the images include: the average colour in the YIQ, HSV, and RGB colour spaces, colour histograms in the HSV and RGB (8 bin and 64 bin) colour spaces a cumulative colour histogram in the HSV colour space, and the moments of colour distribution in the HSV and RGB colour spaces. The texture features extracted from the images include: contrast, coarseness and four variations of directionality. Six distance metrics are used. They are: L 1, L2, weighted L2, Linfinity, weighted sum, and scalar distance.; The objective of this project was to find the most appropriate combinations of features and distance metrics for use in an interior decorating application. (Abstract shortened by UMI.)