The concept, procedure and tools for interactive image retrieval with multiple seed images are investigated in this research. In this paper, we consider integrating the semantic image organization and low level feature representation by user interactivity. First, high level image classification has been initiated by a supervised learning process. The classification result is dynamically adjusted by the feedback information extracted from the interactive query process, which is considered as a long term interactive refinement process. Second, the initial interactive retrieval procedure is determined by both low level features and high level semantic classification. The procedure is further refined by the user feedback so that the meaning of "similarity" defined by a specific user for a particular application can be approached gradually. The process is regarded as short term refinement. With the help of interactive query operation, system performance can never stop to be improved.
[1]
Romain Murenzi,et al.
Fast texture database retrieval using extended fractal features
,
1997,
Electronic Imaging.
[2]
Christos Faloutsos,et al.
QBIC project: querying images by content, using color, texture, and shape
,
1993,
Electronic Imaging.
[3]
C.-C. Jay Kuo,et al.
Terrain texture synthesis with an extended self-similar model
,
1995
.
[4]
William M. Shaw,et al.
Termrelevance Computations and Perfect Retrieval Performance
,
1995,
Inf. Process. Manag..
[5]
John P. Oakley,et al.
Storage and Retrieval for Image and Video Databases
,
1993
.
[6]
Shih-Fu Chang,et al.
Exploring image functionalities in WWW applications development of image/video search and editing engines
,
1997,
Proceedings of International Conference on Image Processing.