Distributions From Images

Slides from a talk outlining the Flexible Histogram method of modeling images. Flexible Histograms capture the texture characteristics within images, using an image representation which measures the joint occurrence of features across spatial resolutions. An image classification system was designed using a similarity metric based on the likelihood that the distribution derived from one image could have generated another. Classification of natural textures indicates a high level of specificity, and recent results on target detection in SAR imagery are encouraging.

Jeremy S. De Bonet ( jsd@ai.mit.edu)

Texture Based Segmentation

A texture driven segmentation built upon the flexible histograms texture matching technique developed in his master's thesis. Examples include segmentation of target vehicles in synthetic aperature radar (SAR), and anatomical structures from magnetic resonance imagery (MRI)

Jeremy S. De Bonet ( jsd@ai.mit.edu)

Toward Automatic Segmentation of SAR Images

In many recognition and classification applications, fast segmentation performed at the pre-processing stage can save a lot of time by clipping out the areas where there could be no objects present. When applied to the SAR domain, segmentation can save not only processing time, but also transmission time and resources. Only the parts of the image that can potentially contain objects of interest are transmitted from a platform to a processing center.

Polina Golland, Paul Viola ( polina@ai.mit.edu)

Structure Driven Image Regisration

Because of its ability to provide a representation which is generally robust to the speckle in synthetic aperatiure radar (SAR) imagery, the flexible histograms texture matching technique developed in his master's thesis can be used as core matching metric for a SAR image registration system. While working on this project during the summer of 1997 at MIT and Alphatech, Inc. he developed such a system.

Jeremy S. De Bonet ( jsd@ai.mit.edu)