Showing papers by "Paul Sajda published in 2000"

Higher-Order Statistical Properties Arising from the Non-Stationarity of Natural Signals

Abstract: We present evidence that several higher-order statistical properties of natural images and signals can be explained by a stochastic model which simply varies scale of an otherwise stationary Gaussian process. We discuss two interesting consequences. The first is that a variety of natural signals can be related through a common model of spherically invariant random processes, which have the attractive property that the joint densities can be constructed from the one dimensional marginal. The second is that in some cases the non-stationarity assumption and only second order methods can be explicitly exploited to find a linear basis that is equivalent to independent components obtained with higher-order methods. This is demonstrated on spectro-temporal components of speech.

Mosaicing and enhancement of images for ophthalmic diagnosis and documentation

Abstract: The invention is a method and system for mosaicing images of the eye to create high resolution, wide-field ophthalmic images for the diagnosis and treatment of eye diseases. The inventive method acquires a first image of the eye, acquires a second image of the eye, and processes the images to produce a mosaic representation. The second image includes a portion of the first image. To guide in acquiring the second image, the method may view the first image while acquiring the second image. The images typically are acquired using either a direct ophthalmoscope or a slitlamp biomicroscope. The method further may convert the images to a digital format. The step of processing includes aligning and merging the images, and conducting real-time processing and non-real-time processing. Real-time processing may include eliminating non-overlapping images, image registration, topology inference, local-to-global alignment, image fusion, signal estimation, and multiresolution blending, while the non-real-time processing may include local alignment, global alignment, image registration, intra-alignment, inter-alignment, signal averaging, and photometric blending.

Hierarchical image probability (HIP) models

Abstract: We formulate a model for probability distributions on image spaces. We show that any distribution of images can be factored exactly into conditional distributions of feature vectors at one resolution (pyramid level) conditioned on the image information at lower resolutions. We would like to factor this over positions in the pyramid levels to make it tractable, but such factoring may miss long-range dependencies. To capture long-range dependencies, we introduce hidden class labels at each pixel in the pyramid. The result is a hierarchical mixture of conditional probabilities, similar to a hidden Markov model on a tree. The model parameters can be found with maximum likelihood estimation using the EM algorithm. We have obtained encouraging preliminary results on the problems of detecting various objects in SAR images and target recognition in optical aerial images.

System and method for providing programming content in response to an audio signal

Abstract: A system and a concomitant method for providing programming content in response to an audio signal. The programming content and the audio signal are transmitted in a network having a forward channel and a back channel. In one embodiment, the system comprises a local processing unit and a remote server computer. A first user provides a first audio signal containing a request for programming content from a service provider. The local processing unit receives the first audio signal and transmits the received first audio signal to a service provider via the back channel. The remote server computer receives the first audio signal from the back channel, recognizes the first user and the request for programming content, retrieves the requested programming content from a program database and transmits the programming content to the local processing unit via the forward channel.

Mammographic mass detection with a hierarchical image probability (HIP) model

Abstract: We formulate a model for probability distributions on image spaces. We show that any distribution of images can be factored exactly into conditional distributions of feature vectors at one resolution (pyramid level) conditioned on the image information at lower resolutions. We would like to factor this over positions in the pyramid levels to make it tractable, but such factoring may miss long-range dependencies. To fix this, we introduce hidden class labels at each pixel in the pyramid. The result is a hierarchical mixture of conditional probabilities, similar to a hidden Markov model on a tree. The model parameters can be found with maximum likelihood estimation using the EM algorithm. We have obtained encouraging preliminary results on the problems of detecting masses in mammograms.

Hierarchical, multi-resolution models for object recognition: applications to mammographic computer-aided diagnosis

Abstract: A fundamental problem in image analysis is the integration of information across scale to detect and classify objects We have developed, within a machine learning framework, two classes of multiresolution models for integrating scale information for object detection and classification-a discriminative model called the hierarchical pyramid neural network and a generative model called a hierarchical image probability model Using receiver operating characteristic analysis, we show that these models can significantly reduce the false positive rates for a well-established computer-aided diagnosis system