Lucent Technologies Inc.

Alcatel Lucent

Agere Systems Inc.

Cadess.ai Ab

Alcatel-lucent USA Inc.

Ingrid Birgitta Carlbom

Yves Darly Jean

Daniel G Aliaga

Thomas A Funkhouser

Agata Opalach

Gopal Sarma Pingali

Sarma Vgk Pingali

Dimah V Yanovsky

James Edward West

Christophe Avenel

Gary Wayne Elko

Man Mohan Sondhi

Sarma V G K Pingali

The method according to the invention utilizes a color decomposition of histological tissue image data to derive a density map. The density map corresponds to the portion of the image data that contains the stain/tissue combination corresponding to the stroma, and at least one gland is extracted from said density map. The glands are obtained by a combination of a mask and a seed for each gland derived by adaptive morphological operations, and the seed is grown to the boundaries of the mask. The method may also derive an epithelial density map used to remove small objects not corresponding to epithelial tissue. The epithelial density map may further be utilized to improve the identification of glandular regions in the stromal density map. The segmented gland is extracted from the tissue data utilizing the grown seed as a mask. The gland is then classified according to its associated features.

Segmentation of histological tissue images into glandular structures for prostate cancer tissue classification

The method of the present invention classifies the nuclei in prostate tissue images with a trained deep learning network and uses said nuclear classification to classify regions, such as glandular regions, according to their malignancy grade. The method according to the present disclosure also trains a deep learning network to identify the category of each nucleus in prostate tissue image data, said category representing the malignancy grade of the tissue surrounding the nuclei. The method of the present disclosure automatically segments the glands and identifies the nuclei in a prostate tissue data set. Said segmented glands are assigned a category by at least one domain expert, and said category is then used to automatically assign a category to each nucleus corresponding to the category of said nucleus' surrounding tissue. A multitude of windows, each said window surrounding a nucleus, comprises the training data for the deep learning network.

Prostate cancer tissue image classification with deep learning

Techniques for computing error-bounded position and orientation (pose) of a panoramic camera in real-world environments. Such environments may include large interior spaces (e.g., buildings) A space may include multiple rooms. For example, a technique for capturing images associated with an environment includes the following steps/operations. First, respective placements of fiducials in the environment are determined so as to satisfy at least one constraint. Images are captured, with an image capture device (e.g., camera), associated with the environment with the fiducials placed therein. A pose estimation of the image capture device is then determined based on projections of the fiducials in the captured images. The pose estimation may be optimized so as to obtain an optimal pose per image. Also, the fiducial placements may be optimized so as to obtain optimal fiducial placements. Then at least one constraint may include a constraint associated with the number of visible fiducials, a constraint associated with a distance from a viewpoint to a fiducial, and/or a constraint associated with an angle subtended by pairs of fiducials.

Method and apparatus for computing error-bounded position and orientation of panoramic cameras in real-world environments

Techniques for intelligent and automatic control of sensors for capturing data associated with real time events. Preferably, the sensors are associated with a multimedia database system. For example, a technique for controlling one or more sensors used to capture data associated with an event comprises the following steps/operations. First, sensor data captured in accordance with the event is processed. Then, the one or more sensors are automatically controlled based on information pertaining to the continual activity of at least one of one or more objects and one or more persons associated with the event in real time obtained using at least a portion of the processed data captured in accordance with the event.

Method and apparatus for intelligent and automatic sensor control using multimedia database system

An acoustic modeling system and an acoustic modeling method use beam tracing techniques that accelerate computation of significant acoustic reverberation paths in a distributed virtual environment. The acoustic modeling system and method perform a priority-driven beam tracing to construct a beam tree data structure representing 'early' reverberation paths between avatar locations by performing a best-first traversal of a cell adjacency graph that represents the virtual environment. To further accelerate reverberation path computations, the acoustic modeling system and method according to one embodiment perform a bi-directional beam tracing algorithm that combines sets of beams traced from pairs of avatar locations to efficiently find viable acoustic reverberation paths.

Acoustic modeling apparatus and method using accelerated beam tracing techniques

A performance data mining system combines detailed sensor analysis data with other data sources to discover interesting patterns/rules for performance and utilizes real time sensor analysis to dynamically derive mining results in real time during an event. The system automatically generates advice/strategy and predictions based on specified criteria.

Performance data mining based on real time analysis of sensor data

Techniques for computing a globally consistent set of image feature correspondences across a wide range of viewpoints suitable for interactive walkthroughs and visualizations. The inventive approach takes advantage of the redundancy inherent in a dense set of images captured in a plane (or in higher dimensions, e.g., images captured in a volume, images captured over time, etc). The technique may detect features in a set of source images and track the features to neighboring images. When features track to the same position in the same image, they are flagged as potential correspondences. Among the potential correspondences, the technique selects the maximal set using a greedy graph-labeling algorithm (e.g., best-first order). Only correspondences that produce a globally consistent labeling are selected. After globalization is done, a set of features common to a group of images can be quickly found and used to warp and combine the images to produce an interpolated novel view of the environment.

Method and apparatus for finding feature correspondences between images captured in real-world environments

Techniques for compressing and decompressing images captured from viewpoints throughout an n-dimensional space. As opposed to conventional approaches, these techniques exploit image coherence over the entire n-dimensional space, providing quick access to images along arbitrary contiguous viewpoint paths during computer graphics applications. This is accomplished by combining a hierarchical data structure with a compression methodology. Such techniques provide for generating and arranging reference images and residual images into a spatial image hierarchy such as a binary tree structure. Images and residuals may be encoded using intra-image compression. A captured image is extracted via a sequence of image warping and addition operations.

Method and apparatus for compressing and decompressing images captured from viewpoints throughout N-dimensional space

Techniques for indexing multimedia data substantially concurrently or contemporaneously with its capture to convert a real world event into an accessible database in real time are provided. The present invention introduces a new paradigm of converting a real world event in real time into a rich multimedia database by processing data from multiple sensors observing the event. Real time analysis of the sensor data, coupled with domain knowledge, results in instant indexing of multimedia data at capture time itself. This yields the semantic information to answer complex queries about the content, and the ability to extract portions of data that correspond to complex actions performed in the real world.

Instantly indexed databases for multimedia content analysis and retrieval

Techniques for retrieval of multimedia data through visual representations are provided. Such visual representations, preferably in the form of visual activity maps or spatio-temporal activity maps, serve as an efficient and intuitive graphical user interface for multimedia retrieval, particularly when the media streams are derived from multiple sensors observing a physical environment. An architecture for interactive media retrieval is also provided by combining such visual activity maps with domain specific event information. Visual activity maps are derived from the trajectories of motion of objects in the environment. The visual activity map based techniques significantly help users in quickly and effectively discovering interesting portions of the data, and randomly accessing and retrieving the corresponding portions of the media streams.

Method and apparatus for retrieving multimedia data through spatio-temporal activity maps

Techniques for creating interactive walkthroughs or visualizations of real-world environments. Such environments for which interactive walkthroughs may be created may comprise large interior spaces (e.g., buildings), wherein a space may comprise multiple rooms. The interactive walkthroughs may be photorealistic and rendered from arbitrary viewpoints at interactive rates. In one aspect of the invention, a technique for generating a visualization representative of an environment comprises the following steps. A dense set of image data is captured from the environment. The captured image data is compressed for storage. A portion of the compressed image data may then be decompressed so as to render a visualization representative of the environment from a selected novel viewpoint. Another portion of the image data may be prefetched and decompressed such that captured image data surrounding the selected novel viewpoint is available to be rendered. The visualization of the environment may then be rendered from the selected novel viewpoint.

Method and system for creating interactive walkthroughs of real-world environment from set of densely captured images

A method and apparatus for tracking an object using one or more video cameras together with a plurality of microphones and/or geophones, whereby accurate tracking of the object can be obtained even when the object changes directions as a result of an impact. A tennis ball may be tracked during a tennis match. Prior to and after an impact, the ball is tracked with the use of video cameras, illustratively in accordance with the method disclosed in U.S. Pat. No. 6,233,007. Then, the position of the ball at racket or ground impact is accurately determined by measuring the acoustical waves generated by the ball hitting the racket and/or either the acoustic waves or the elastic (Rayleigh) waves generated by the ball hitting the ground, respectively. Tracking then advantageously continues with use of the video cameras.

Real-time method and apparatus for tracking a moving object experiencing a change in direction

An omnidirectional video camera captures images of the environment while moving along several intersecting paths forming an irregular grid. These paths define the boundaries of a set of image loops within the environment. For arbitrary viewpoints within each image loop, a 4D plenoptic function may be reconstructed from the group of images captured at the loop boundary. For an observer viewpoint, a strip of pixels is extracted from an image in the loop in front of the observer and paired with a strip of pixels extracted from another image on the opposite side of the image loop. A new image is generated for an observer viewpoint by warping pairs of such strips of pixels according to the 4D plenoptic function, blending each pair, and then stitching the resulting strips of pixels together.

Growing community of inventors

Summit, NJ, United States of America

Ingrid Birgitta Carlbom