Blobworld is a system for image retrieval that focuses on finding coherent image regions, or "blobs," which correspond to objects or parts of objects. Each image is automatically segmented into these regions, each with associated color and texture descriptors. Querying is based on the attributes of one or two regions of interest rather than the entire image. To enable large-scale retrieval, the blob descriptions are indexed using a tree structure, and a lower-rank approximation is used to handle the high-dimensional feature space. Experiments show that Blobworld outperforms traditional methods in terms of query precision and the ease of refining queries. The system is evaluated using a collection of 10,000 Corel images, demonstrating higher precision for querying distinctive objects like tigers, zebras, and cheetahs compared to querying using color and texture histograms of the entire image. Additionally, Blobworld's false positives are easier to understand due to the highlighting of matching regions. The paper also discusses an indexing approach that projects color feature vectors to a lower-dimensional space, achieving query results of similar quality to those from scanning the entire database.Blobworld is a system for image retrieval that focuses on finding coherent image regions, or "blobs," which correspond to objects or parts of objects. Each image is automatically segmented into these regions, each with associated color and texture descriptors. Querying is based on the attributes of one or two regions of interest rather than the entire image. To enable large-scale retrieval, the blob descriptions are indexed using a tree structure, and a lower-rank approximation is used to handle the high-dimensional feature space. Experiments show that Blobworld outperforms traditional methods in terms of query precision and the ease of refining queries. The system is evaluated using a collection of 10,000 Corel images, demonstrating higher precision for querying distinctive objects like tigers, zebras, and cheetahs compared to querying using color and texture histograms of the entire image. Additionally, Blobworld's false positives are easier to understand due to the highlighting of matching regions. The paper also discusses an indexing approach that projects color feature vectors to a lower-dimensional space, achieving query results of similar quality to those from scanning the entire database.