mechanisms

Segmenting 3D textured surfaces is critical for general image understanding.

Image texture is useful for segmentation and for computing surface orientations of uniformly textured objects. If texture is ignored, it can cause failure for stereo and gray-scale segmentation algorithms. In the past, mathematical representations of image texture have been applied to only specific texture problems, and no consideration has been given to the models' generality across different computer vision tasks and different image phenomena. We advocate the space/frequency representation, which shows the local spatial frequency content of every point in the image. From several different methods of computing the representation, we pick the spectrogram. The spectrogram elucidates many disparate image phenomena including texture boundaries, texture in perspective, aliasing, zoom, and blur. Many past

Current image retrieval techniques, while extremely accurate and effective on a relatively small scale, are unable to scale to very large database conditions due to the limitations imposed by the curse of dimensionality on nearest neighbor searching. In addition to this, the better systems are unsuited to visual browsing of the database. We propose the HUGINN framework for a concept-based image retrieval system, which uses state of the art machine learning techniques to constrain searching and browsing through the use of image concept. This report outlines the current state of the art in image retrieval and outlines a current system using these techniques and a novel technique called probabilistic pose prediction to achieve excellent results for near-duplicate and subimage retrieval. It then details a novel system for boosting machine learners to aid in the extraction of concept from images, then outlines a plan for future research culminating in a complete image retrieval and browsing system which incorporates machine learning.