Abstract: One embodiment of the present invention sets forth a technique, which includes dividing an input image into a first partial image that stores a first subset of bits in each pixel of the input image and a second partial image that stores a second subset of bits that is disjoint from the first subset of bits in each pixel of the input image. The technique also includes modifying a first set of pixels in the first partial image to generate a first partial image processing result and modifying a second set of pixels in the second partial image to generate a second partial image processing result. The technique further includes generating a combined image processing result based on a combination of the first partial image processing result and the second partial image processing result.

Abstract: The disclosed computer-implemented method may include generating a three-dimensional (3D) feature map for a digital image using a fully convolutional network (FCN). The 3D feature map may be configured to identify features of the digital image and identify an image region for each identified feature. The method may also include generating a region composition graph that includes the identified features and image regions. The region composition graph may be configured to model mutual dependencies between features of the 3D feature map. The method may further include performing a graph convolution on the region composition graph to determine a feature aesthetic value for each node according to the weightings in the node's weighted connecting segments, and calculating a weighted average for each node's feature aesthetic value to provide a combined level of aesthetic appeal for the digital image. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: In various embodiments, a defective pixel detection application automatically detects defective pixels in video content. In operation, the defective pixel detection application computes a first set of pixel intensity gradients based on a first frame of video content and a first neighborhood of pixels associated with a first pixel. The defective pixel detection application also computes a second set of pixel intensity gradients based on the first frame and a second neighborhood of pixels associated with the first pixel. Subsequently, the defective pixel detection application computes a statistical distance between the first set of pixel intensity gradients and the second set of pixel intensity gradients. The defective pixel detection application then determines that the first pixel is defective based on the statistical distance.

Abstract: The disclosed computer-implemented method may include generating a three-dimensional (3D) feature map for a digital image using a fully convolutional network (FCN). The 3D feature map may be configured to identify features of the digital image and identify an image region for each identified feature. The method may also include generating a region composition graph that includes the identified features and image regions. The region composition graph may be configured to model mutual dependencies between features of the 3D feature map. The method may further include performing a graph convolution on the region composition graph to determine a feature aesthetic value for each node according to the weightings in the node's weighted connecting segments, and calculating a weighted average for each node's feature aesthetic value to provide a combined level of aesthetic appeal for the digital image. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: A method and system for classification of endoscopic images is disclosed. An initial trained deep network classifier is used to classify endoscopic images and determine confidence scores for the endoscopic images. The confidence score for each endoscopic image classified by the initial trained deep network classifier is compared to a learned confidence threshold. For endoscopic images with confidence scores higher than the learned threshold value, the classification result from the initial trained deep network classifier is output. Endoscopic images with confidence scores lower than the learned confidence threshold are classified using a first specialized network classifier built on a feature space of the initial trained deep network classifier.

Abstract: In various embodiments, a defective pixel detection application automatically detects defective pixels in video content. In operation, the defective pixel detection application computes a first set of pixel intensity gradients based on a first frame of video content and a first neighborhood of pixels associated with a first pixel. The defective pixel detection application also computes a second set of pixel intensity gradients based on the first frame and a second neighborhood of pixels associated with the first pixel. Subsequently, the defective pixel detection application computes a statistical distance between the first set of pixel intensity gradients and the second set of pixel intensity gradients. The defective pixel detection application then determines that the first pixel is defective based on the statistical distance.

Abstract: A method and system for classification of endoscopic images is disclosed. An initial trained deep network classifier is used to classify endoscopic images and determine confidence scores for the endoscopic images. The confidence score for each endoscopic image classified by the initial trained deep network classifier is compared to a learned confidence threshold. For endoscopic images with confidence scores higher than the learned threshold value, the classification result from the initial trained deep network classifier is output. Endoscopic images with confidence scores lower than the learned confidence threshold are classified using a first specialized network classifier built on a feature space of the initial trained deep network classifier.

Abstract: A method for performing cellular classification includes generating a plurality of local dense Scale Invariant Feature Transform (SIFT) features based on a set of input images and converting the plurality of local dense SIFT features into a multi-dimensional code using a feature coding process. A first classification component is used to generate first output confidence values based on the multi-dimensional code and a plurality of global Local Binary Pattern Histogram (LBP-H) features are generated based on the set of input images. A second classification component is used to generate second output confidence values based on the plurality of LBP-H features and the first output confidence values and the second output confidence values are merged. Each of the set of input images may then be classified as one of a plurality of cell types using the merged output confidence values.

Abstract: Independent subspace analysis (ISA) is used to learn (42) filter kernels for CLE images in brain tumor classification. Convolution (46) and stacking are used for unsupervised learning (44, 48) with ISA to derive the filter kernels. A classifier is trained (56) to classify CLE brain images based on features extracted using the filter kernels. The resulting filter kernels and trained classifier are used (60, 64) to assist in diagnosis of occurrence of brain tumors during or as part of neurosurgical resection. The classification may assist a physician in detecting whether CLE examined brain tissue is healthy or not and/or a type of tumor.

Abstract: A method and system for classifying tissue endomicroscopy images are disclosed. Local feature descriptors are extracted from an endomicroscopy image. Each of the local feature descriptors is encoded using a learnt discriminative dictionary. The learnt discriminative dictionary includes class-specific sub-dictionaries and penalizes correlation between bases of sub-dictionaries associated with different classes. Tissue in the endomicroscopy image is classified using a trained machine learning based classifier based on the coded local feature descriptors encoded using a learnt discriminative dictionary.

Abstract: A method for performing cellular classification includes generating a plurality of local dense Scale Invariant Feature Transform (SIFT) features based on a set of input images and converting the plurality of local dense SIFT features into a multi-dimensional code using a feature coding process. A first classification component is used to generate first output confidence values based on the multi-dimensional code and a plurality of global Local Binary Pattern Histogram (LBP-H) features are generated based on the set of input images. A second classification component is used to generate second output confidence values based on the plurality of LBP-H features and the first output confidence values and the second output confidence values are merged. Each of the set of input images may then be classified as one of a plurality of cell types using the merged output confidence values.

Abstract: A method for transmitting a video data over a grid infrastructure network is disclosed. The method includes receiving a request from at least one user for viewing the video data and identifying a plurality of attributes from a plurality primary grid enabled mini servers (GEMS), wherein the plurality of primary GEMS together form the grid network. The method further includes partitioning video data into a plurality of discrete fragments using a shard creator indicative of the plurality of attributes in each of the plurality of primary GEMS and allocating the plurality of discrete fragments among the plurality of primary GEMS based on the plurality of attributes of each of the plurality of primary GEMS. The method also includes decoding the plurality of discrete fragments of the video data using a streaming server for transmitting the video data to the at least one user.

Abstract: A method for transmitting a video data over a grid infrastructure network is disclosed. The method includes receiving a request from at least one user for viewing the video data and identifying a plurality of attributes from a plurality primary grid enabled mini servers (GEMS), wherein the plurality of primary GEMS together form the grid network. The method further includes partitioning video data into a plurality of discrete fragments using a shard creator indicative of the plurality of attributes in each of the plurality of primary GEMS and allocating the plurality of discrete fragments among the plurality of primary GEMS based on the plurality of attributes of each of the plurality of primary GEMS. The method also includes decoding the plurality of discrete fragments of the video data using a streaming server for transmitting the video data to the at least one user.