In the past year, the research for content-based image retrieval has been focused on the following issues:
The team has developed a pattern thesaurus for characterizing the texture information in large aerial photographs [5]
The salient components of this system include texture feature extraction, image segmentation and grouping, learning similarity measure, and a texture thesaurus model for fast search and indexing. The texture features are computed by filtering the image with a bank of Gabor filters. This is followed by a texture gradient computation to segment each large airphoto into homogeneous regions. A hybrid neural network algorithm is used to learn the visual similarity by clustering patterns in the feature space. With learning similarity, the retrieval performance improves significantly[8]. Finally, a texture image thesaurus is created by combining the learning similarity algorithm and a hierarchical vector quantization scheme. This thesaurus facilitates the indexing process while maintaining a good retrieval performance. Experimental results demonstrate the robustness of the overall system in searching over a large collection of aerial photographs.
A texture thesaurus can be visualized as an image counterpart of the traditional thesaurus for text search. Similar to parsing text documents using a dictionary, the information within images can be classified and indexed using a texture thesaurus. A similarity learning algorithm is used to combine human perceptual similarity with the low level texture feature vectors to obtain clusters in the feature space. This is followed by a hierarchical vector quantizer to construct the texture code words. The hierarchical quantization provides, in addition, an efficient indexing tree while maintaining or even improving the similarity retrieval. The visual code word representation in the thesaurus can be used as information samples to help users browse through the database. More details can be found in [5]. Figure 4 show some retrieval examples from a demonstration program.
Figure 4: Two Examples of Retrieval.
In order to make the individual object or region of the images indexable, the input images have to be partitioned into homogeneous regions at the time of ingest into the database. We have proposed an edge flow model for general purpose boundary detection and image segmentation, and also implemented a tool which provides users an easy way to process large and diverse collections of images [3, 4].
This scheme utilizes a predictive coding model to identify the direction of change in color and texture at each image location at a given scale, and constructs an edge flow vector. By iteratively propagating the edge flow, the boundaries can be detected at image locations which encounter two opposite directions of flow in the stable state. A user defined image scale is the only significant control parameter that is needed by the algorithm. The scheme also facilitates integration of color, texture, and phase into a single framework for boundary detection. Experimental results on a diverse collection of over 2,500 images from a stock photo library have yielded consistently good (visually acceptable) segmentation results. Initial results on image retrieval that combines segmentation with feature extraction have yielded very encouraging results. Figure 5 shows some examples of segmentation/retrieval.
Figure 5: Two Examples of Segmentation and Retrieval.