الفهرس | Only 14 pages are availabe for public view |
Abstract Due to the enormous increase in image database sizes, as well as its vast deployment in various applications, query by content or Content Based Image Retrieval (CBIR) has recently been proposed as an alternative to text-based retrieval for media such as images, videos, and audios. The main problems involved in text-based image retrieval include: keyword annotation is labor intensive so it is hard to index large sets of images using these annotations. Also, these annotations are drawn from a predefined set of keywords, which cannot cover all possible concepts images may represent in addition that keywords assignment is subjective to the person making it. To overcome these problems, content-based image retrieval systems propose to index the media documents based on features extracted from their content rather than by textual annotations. In content-based image retrieval, image data representation and similarity measurement are two important tasks. So, when building CBIR systems, this requires choosing and representing the visual features and finding combinations of techniques that give the best matches and enhance the performance of CBIR systems. This study aims to improve the performance of content-based image retrieval systems via multiple features representations. One way of doing this important task is to extract the visual features from the database images and index them based on these features, then examine the features extraction algorithms and find good combinations of these algorithms by measuring the retrieval accuracy when using them together On the other hand, the choice of the similarity distance measures is also important as the retrieval process will be done based on a comparison between the features vectors of the query image and the corresponding ones of each image in the database. Using different categories of images (e.g. roses, people, beach, and buses), the experiments confirmed the importance of using the spatial information beside the color feature itself and showed that: Haar and Daubechies wavelets can be combined with Global Color Histogram (GCH) and the color layout feature for efficient content-based image retrieval systems. Also, the best retrieval accuracy is obtained when combining the Haar wavelet with GCH in addition to the color layout feature using the Euclidean distance measure, while also the cosine similarity measure gives good results if compared with the other used similarity measures like Manhattan and correlation similarity measures. The work presented here can be generalized and used in. |