Abstract: The present disclosure relates to encoding and decoding of a picture or a plurality of pictures (e.g. video) using a neural network which is partially trained online. Accordingly, at an encoder one or more layers are selected which are to be trained. Then, the training of the neural network is performed in which parameters of the selected layers are updated. The parameters of the remaining layers are maintained and not updated. The updated parameters are provided within a bitstream. The picture(s) is/are also encoded. The decoder receives an indication of the updated parameters, updates only those parameters indicated, and applies the so obtained neural network.

Abstract: A method of similarity-based deduplication comprising the steps of: receiving an input data block; computing discrete wavelet transform (DWT) coefficients; extracting feature-related DWT data from the computed DWT coefficients; applying quantization to the extracted feature-related DWT data to obtain keys as results of the quantization; constructing a locality-sensitive fingerprint of the input data block; computing a similarity degree between the locality-sensitive fingerprint of the input data block and a locality-sensitive fingerprint of each data block in the plurality of the data blocks in a cache memory; selecting an optimal reference data block as the data block; determining a differential compression is required to be applied based on the similarity degree between the input data block and the optimal reference data block; applying the differential compression to the input data block and the optimal reference data block.

Abstract: A method of similarity-based deduplication comprising the steps of: receiving an input data block; computing discrete wavelet transform (DWT) coefficients; extracting feature-related DWT data from the computed DWT coefficients; applying quantization to the extracted feature-related DWT data to obtain keys as results of the quantization; constructing a locality-sensitive fingerprint of the input data block; computing a similarity degree between the locality-sensitive fingerprint of the input data block and a locality-sensitive fingerprint of each data block in the plurality of the data blocks in a cache memory; selecting an optimal reference data block as the data block; determining a differential compression is required to be applied based on the similarity degree between the input data block and the optimal reference data block; applying the differential compression to the input data block and the optimal reference data block.

Abstract: Systems and methods for image processing may perform one or more operations including, but not limited to: receiving raw image data from at least one imaging device; computing at least one image depth distance from the raw image data; computing one or more image validity flags from the raw image data; generating at least one data validity mask from the one or more image validity flags; determining a background imagery estimation from at least one image depth distance; generating at least one foreground mask from the background imagery estimation and the at least one image depth distance; generating at least one region-of-interest mask from the data validity mask and the foreground mask; and generating filtered raw image data from the raw image data and at least one region of interest mask.

Abstract: An image processing system comprises an image processor having image processing circuitry and an associated memory. The image processor is configured to implement a foreground processing module utilizing the image processing circuitry and the memory. The foreground processing module is configured to obtain one or more images, to estimate a foreground region of interest from the one or more images, to determine a plurality of segments of the foreground region of interest, to calculate amplitude statistics for respective ones of the plurality of segments, to classify respective segments as being respective portions of static foreground objects or as being respective portions of dynamic foreground objects based at least in part on the calculated amplitude statistics and one or more defined patterns for known static and dynamic objects, and to remove one or more segments classified as static foreground objects from the foreground region of interest.

Abstract: A method and system for key frame based region of interest (ROI) tracking is disclosed. The method includes storing a key ROI set in a key ROI buffer, the key ROI set including at least one key ROI; designating one of the key ROI in the key ROI set as an active key ROI; receiving a point cloud representing a particular ROI to be processed for tracking; establishing a correspondence between that particular ROI and the active key ROI; determining whether to switch the active key designation to another key ROI in the key ROI set and switching the active key designation accordingly; and determining whether to modify the key ROI set and modifying the key ROI set accordingly.

Abstract: In one embodiment, an image processor comprises image processing circuitry and an associated memory. The image processor is configured to implement a gesture recognition system utilizing the image processing circuitry and the memory. The gesture recognition system implemented by the image processor comprises a dynamic gesture recognition module. The dynamic gesture recognition module is configured to establish a dynamic gesture recognition interval comprising a plurality of image frames, to extract one or more first features from the dynamic gesture recognition interval, to adjust the dynamic gesture recognition interval, to extract one or more second features from the adjusted dynamic gesture recognition interval, and to recognize a dynamic gesture based at least in part on at least a subset of the extracted first and second features.

Abstract: Systems and methods for image processing may perform one or more operations including, but not limited to: receiving raw image data from at least one imaging device; computing at least one image depth distance from the raw image data; computing one or more image validity flags from the raw image data; generating at least one data validity mask from the one or more image validity flags; determining a background imagery estimation from at least one image depth distance; generating at least one foreground mask from the background imagery estimation and the at least one image depth distance; generating at least one region-of-interest mask from the data validity mask and the foreground mask; and generating filtered raw image data from the raw image data and at least one region of interest mask.

Abstract: An image processing system comprises an image processor having image processing circuitry and an associated memory. The image processor is configured to implement an object tracking module. The object tracking module is configured to obtain one or more images, to extract contours of at least two objects in at least one of the images, to select respective subsets of points of the contours for the at least two objects based at least in part on curvatures of the respective contours, to calculate features of the subsets of points of the contours for the at least two objects, to detect intersection of the at least two objects in a given image, and to track the at least two objects in the given image based at least in part on the calculated features responsive to detecting intersection of the at least two objects in the given image.

Abstract: An image processing system comprises an image processor having image processing circuitry and an associated memory. The image processor is configured to implement a face recognition system utilizing the image processing circuitry and the memory, the face recognition system comprising a face recognition module. The face recognition module is configured to identify a region of interest in each of two or more images, to extract a three-dimensional representation of a head from each of the identified regions of interest, to transform the three-dimensional representations of the head into respective two-dimensional grids, to apply temporal smoothing to the two-dimensional grids to obtain a smoothed two-dimensional grid, and to recognize a face based on a comparison of the smoothed two-dimensional grid and one or more face patterns.

Abstract: An image processing system comprises an image processor having image processing circuitry and an associated memory. The image processor is configured to implement a foreground processing module utilizing the image processing circuitry and the memory. The foreground processing module is configured to obtain one or more images, to estimate a foreground region of interest from the one or more images, to determine a plurality of segments of the foreground region of interest, to calculate amplitude statistics for respective ones of the plurality of segments, to classify respective segments as being respective portions of static foreground objects or as being respective portions of dynamic foreground objects based at least in part on the calculated amplitude statistics and one or more defined patterns for known static and dynamic objects, and to remove one or more segments classified as static foreground objects from the foreground region of interest.

Abstract: An image processing system comprises an image processor having image processing circuitry and an associated memory. The image processor is configured to implement a gesture recognition system comprising a contour classification module. The contour classification module is configured to identify one or more hand poses from one or more isolated regions in a first image, to determine a contour of a given one of the one or more hand poses, to calculate one or more features of the contour of the given hand pose, to identify one or more isolated regions in a second image, and to determine whether at least a portion of one or more isolated regions in the second image matches the given hand pose based on a comparison of one or more points characterizing the portion of the one or more isolated regions in the second image and the one or more features of the contour of the given hand pose.

Abstract: An image processing system comprises an image processor having image processing circuitry and an associated memory. The image processor is configured to implement a gesture recognition system utilizing the image processing circuitry and the memory. The gesture recognition system comprises a finger detection and tracking module configured to identify a hand region of interest in a given image, to extract a contour of the hand region of interest, to detect fingertip positions using the extracted contour, and to track movement of the fingertip positions over multiple images including the given image.

Abstract: A machine-implemented method of generating trapping-set information for use in LDPC-decoding processing of read signals generated, e.g., by sensing a storage medium, such as a magnetic platter. In one embodiment, the method can be implemented as an add-on to any other trapping-set search method in which the discovered trapping sets are evaluated to determine their influence on the overall bit-error rate and/or error-floor characteristics of the LDPC decoder. The method can advantageously reuse at least some of the computational results obtained during this evaluation, thereby requiring a relatively small amount of additional computations, while providing a significant benefit of discovering many more trapping sets in addition to the ones that are being evaluated.

Abstract: A method and system for registration of three-dimensional (3D) image frames is disclosed. The method includes receiving two point clouds representing two 3D image frames obtained at two time instances; locating the origins for the two point clouds; constructing two 2D grids for representing the two point clouds, wherein each 2D grid is constructed based on spherical representation of its corresponding point cloud and origin; identifying two sets of feature points based on the two 2D grids constructed; establishing a correspondence between the first set of feature points and the second set of feature points based on a neighborhood radius threshold; and determining an orthogonal transformation between the first 3D image frame and the second 3D image frame based on the correspondence between the first set of feature points and the second set of feature points.

Abstract: A method and system for key frame based region of interest (ROI) tracking is disclosed. The method includes storing a key ROI set in a key ROI buffer, the key ROI set including at least one key ROI; designating one of the key ROI in the key ROI set as an active key ROI; receiving a point cloud representing a particular ROI to be processed for tracking; establishing a correspondence between that particular ROI and the active key ROI; determining whether to switch the active key designation to another key ROI in the key ROI set and switching the active key designation accordingly; and determining whether to modify the key ROI set and modifying the key ROI set accordingly.

Abstract: A video transcoder for converting a compressed input video bit-stream having one spatial resolution into a compressed output video bit-stream having a different spatial resolution in a manner that enables the transcoder to dynamically change the amount of computational resources allocated to the conversion process. In one embodiment, the video transcoder has a plurality of configurable processing paths whose configuration determines the amount of allocated computational resources. Exemplary processing-path configuration changes may include, but are not limited to engaging or disengaging a processing path, redirecting a data flow from flowing through one processing path to flowing through another processing path, and attaching or detaching one or more processing modules to an engaged processing path.

Abstract: A machine-implemented method of generating trapping-set information for use in LDPC-decoding processing of read signals generated, e.g., by sensing a storage medium, such as a magnetic platter. In one embodiment, the method can be implemented as an add-on to any other trapping-set search method in which the discovered trapping sets are evaluated to determine their influence on the overall bit-error rate and/or error-floor characteristics of the LDPC decoder. The method can advantageously reuse at least some of the computational results obtained during this evaluation, thereby requiring a relatively small amount of additional computations, while providing a significant benefit of discovering many more trapping sets in addition to the ones that are being evaluated.

Abstract: A computer-aided design method for developing, simulating, and testing a read-channel architecture to be implemented in a VLSI circuit. The method uses a coset operating mode and nonzero-syndrome-based decoding to accelerate the simulation of the read-channel's error-rate characteristics corresponding to different parity-check matrices employed in the read-channel's turbo-decoder, such as a low-density parity-check decoder. The acceleration is achieved through recycling some previously generated log-likelihood-ratio values, which enables the method to sometimes bypass certain time-consuming processing steps therein.

Abstract: A computer-aided design method for developing, simulating, and testing a read-channel architecture to be implemented in a VLSI circuit. The method uses codeword/waveform classification to accelerate simulation of the read-channel's error-rate characteristics, with said classification being generated using a first read-channel simulator having a limited functionality. A second read-channel simulator having an extended functionality is then run only for some of the codewords, with the latter having been identified based on said codeword/waveform classification. The acceleration is achieved, at least in part, because the relatively highly time-consuming processing steps implemented in the second read-channel simulator are applied to fewer codewords than otherwise required by conventional simulation methods.