• Key Words
• Creator's name
• Creator's culture
• Material used
• Year


• Image Content
• Global Color
  We compute the color histogram in HSV space and use histogram intersection method to compute the similarity distance.
• Global Texture
  We compute the three texture features for each image, i.e. Coarseness, Contrast, and Directionality. We use histogram intersection and Euclidean distance to calculate the similarity distance.
• Color/Texture Layout
  We divide each image into 5 by 5 sub-images and compute the color/texture features for each sub-image separately. We have a color picker and texture picker in our demo so the use can fill in the 5 by 5 grid using these color/texture feature and then make a query based on this color/texture layout information.
• Shape
  We first use Attraction-based grouping to segment out the object in the image. Then, we use a modified Fourier Descriptor to represent the object boundary. Since this MFD is invariant to translation, rotation, scaling and starting point of the boundary, the matching is extremely fast.

The image collection comes from the Fowler Museum and is a collection of ancient African artifacts.

• Sharad Mehrotra, Yong Rui, Michael Ortega-Binderberger and Thomas S. Huang, "Supporting Content-based Queries over Images in MARS", submitted to ICMC 97.
• Thomas S. Huang, Sharad Mehrotra and Kannan Ramchandran, "Some Challenging Issues in Image Content-based Access and Retrieval", 33rd Annual Clinic on Library Application of Data Processing "Digital Image Access and Retrieval", March 1996.
• Yong Rui, Alfred C. She, and Thomas S. Huang, "Automated Shape Segmentation Using Attraction-based Groupingc in Spatial-Color-Texture Space", ICIP-96, Lausanne, Switzerland.
• Yong Rui, Alfred C. She, and Thomas S. Huang, "Modified Fourier Descriptors for Shape Representation - A Practical Approach", First International Workshop on Image Databases and Multi Media Search, 1996, Amsterdam, The Netherlands.