Embodiments of the present invention are directed to a computer-implemented method for action localization. A non-limiting example of the computer-implemented method includes receiving, by a processor, a video and segmenting, by the processor, the video into a set of video segments. The computer-implemented method classifies, by the processor, each video segment into a class and calculates, by the processor, importance scores for each video segment of a class within the set of video segments. The computer-implemented method determines, by the processor, a winning video segment of the class within the set of video segments based on the importance scores for each video segment within the class, stores, by the processor, the winning video segment from the set of video segments, and removes the winning video segment from the set of video segments.

A method for deblurring a target image includes receiving a burst of images that includes the target image; partitioning respective images of the burst of images into respective patches; converting, to a frequency domain, the respective patches into respective transform patches; selecting a first set of corresponding transform patches from the respective transform patches, where the first set of the corresponding transform patches includes a respective transform patch for a respective image of the burst of images; obtaining, using a neural network, respective weight maps for the corresponding transform patches; obtaining a deblurred transform patch by combining the first set of corresponding transform patches using the respective weight maps; obtaining a first deblurred patch by converting the deblurred transform patch to a pixel domain; and obtaining a deblurred image of the target image using the first deblurred patch.

A method is provided for measuring or estimating stress distributions on heart valve leaflets by obtaining three-dimensional images of the heart valve leaflets, segmenting the heart valve leaflets in the three-dimensional images by capturing locally varying thicknesses of the heart valve leaflets in three-dimensional image data to generate an image-derived patient-specific model of the heart valve leaflets, and applying the image-derived patient-specific model of the heart valve leaflets to a finite element analysis (FEA) algorithm to estimate stresses on the heart valve leaflets. The images of the heart valve leaflets may be obtained using real-time 3D transesophageal echocardiography (rt-3DTEE). Volumetric images of the mitral valve at mid-systole may be analyzed by user-initialized segmentation and 3D deformable modeling with continuous medial representation to obtain, a compact representation of shape. The regional leaflet stress distributions may be predicted in normal and diseased (regurgitant) mitral valves using the techniques of the invention.

Disclosed are methods and apparatuses to recognize actors during normal system operation. The method includes defining actor input such as hand gestures, executing and detecting input, and identifying salient features of the actor therein. A model is defined from salient features, and a data set of salient features and/or model are retained, and may be used to identify actors for other inputs. A command such as “unlock” may be executed in response to actor input. Parameters may be applied to further define where, when, how, etc. actor input is executed, such as defining a region for a gesture. The apparatus includes a processor and sensor, the processor defining actor input, identifying salient features, defining a model therefrom, and retaining a data set. A display may also be used to show actor input, a defined region, relevant information, and/or an environment. A stylus or other non-human actor may be used.

Disclosed systems and methods generate filtered, three-dimensional models with regard to a site. In particular, one or more embodiments include systems and methods that generate a filtered, three-dimensional model by removing one or more non-ground objects from a three-dimensional model of the site. Specifically, one or more embodiments of the disclosed systems and methods remove objects from a three-dimensional representation of a site by applying an initial filter to the three-dimensional representation, identifying regions corresponding to types of terrain within the three-dimensional representation, and applying another filter with parameters particular to the identified regions.

A computer implemented method to present digital images may include storing a digital image in a database and applying a digital image processing technique to the digital image to identify a region of interest of the digital image. The method may also include storing region data that identifies the region of interest of the digital image in the database and receiving a request for information associated with the digital image from a digital device. In response to the request, the method may include providing the digital image and the region data for transmission to the digital device, the digital device configured to adjust a cropping view of the digital image based on the region data to display the region of interest of the digital image.

An embodiment of a segmentation solution for use in diagnostic imaging applications is proposed. A corresponding embodiment of a data-processing segmentation method comprises: providing a representation over a non-zero analysis time period of a body-part being perfused with a contrast agent, the representation comprising, for each location of a set of locations of the body-part, an indication of a response over the analysis time period of the location to an interrogation signal; calculating, for each selected location of a set of selected locations, the value of at least one statistical parameter of a statistical distribution of the response over the analysis time period of the selected location, the set of selected locations comprising all the locations or a part thereof; and segmenting the selected locations according to a comparison between the values of said at least one statistical parameter for the selected locations with at least one segmentation threshold.

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating a plurality of possible segmentations of an image. In one aspect, a method comprises: receiving a request to generate a plurality of possible segmentations of an image; sampling a plurality of latent variables from a latent space, wherein each latent variable is sampled from the latent space in accordance with a respective probability distribution over the latent space that is determined based on the image; generating a plurality of possible segmentations of the image, comprising, for each latent variable, processing the image and the latent variable using a segmentation neural network having a plurality of segmentation neural network parameters to generate the possible segmentation of the image; and providing the plurality of possible segmentations of the image in response to the request.

The disclosure relates to a method and to an image processing facility configured to carry out the method for the segmentation of image data of a target object. In the method, a first segmentation is generated by a trained algorithm. Furthermore, a statistical shape and appearance model is provided, which is trained on corresponding target objects. An interference region is further determined, in which the image data is impaired by an image artifact. A final segmentation of the image data is then generated by adjusting the shape and appearance model to the respective target object outside the interference region and using the first segmentation in the interference region.

A system and method for multi-spectrum segmentation for computer vision is described. A first sensor captures an image within a first spectrum range and generates first sensor data. A second sensor captures an image within a second spectrum range different than the first spectrum range and generates second sensor data. A multi-spectrum segmentation module identifies a segmented portion of the image within the second spectrum range based on: the second sensor data, a subset of the first sensor data corresponding to the segmented portion of the image within the second spectrum range, and a segmented portion of the image at the second spectrum range corresponding to the subset of the first sensor data. The multi-spectrum segmentation module identifies a physical object in the segmented portion of the image within the second spectrum range, and a device generates augmented reality content based on the identified physical object.