Abstract: Given an input image, an image enhancement task, and no external examples available to train on, an Image-Specific Deep Network is constructed tailored to solve the task for this specific image. Since there are no external examples available to train on, the network is trained on examples extracted directly from the input image itself. The current solution solves the problem of Super-Resolution (SR), whereas the framework is more general and is not restricted to SR.

Abstract: Given an input image, an image enhancement task, and no external examples available to train on, an Image-Specific Deep Network is constructed tailored to solve the task for this specific image. Since there are no external examples available to train on, the network is trained on examples extracted directly from the input image itself. The current solution solves the problem of Super-Resolution (SR), whereas the framework is more general and is not restricted to SR.

Abstract: Devices, systems, and methods of blind deblurring and blind super-resolution utilizing internal patch recurrence. Small signal patches tend to repeat “as is” across multiple scales of a natural signal. This fractal-like behavior is utilized for signal processing tasks, including “Blind Deblurring” or “Blind Super-Resolution”, namely, removing signal blur or increasing signal resolution without a-priori knowledge of the underlying blur kernel. While the cross-scale patch recurrence is strong in signals taken under ideal conditions, the cross-scale patch recurrence significantly diminishes when the acquisition blur deviates from an ideal blur. These deviations from ideal patch recurrences are used for recovering the underlying (unknown) blur kernel. The correct blur kernel is recovered by seeking the kernel which maximizes the patch similarity across scales of a related “reference” signal.

Abstract: Device, system, and method of visual inference by collaborative composition. A method of searching for similar target content-items for a plurality of source content-items, the method implementable on a computing device, comprises: for each of said plurality of source content-items, searching for similar target content-items in a dataset, while taking into account collaborative feedback exchanged among at least some of currently-searched said plurality of source content-items.

Abstract: A method implementable on a computing device includes exploiting data redundancy to combine high frequency information from at least two different scales of an input signal to generate a super resolution version of said input signal. An alternative method includes exploiting recurrence of data from an input signal in at least two different scales of at least one reference signal to extract and to combine high frequency information from a plurality of scales of said at least one reference signal to generate a super resolution version of said input signal. An alternative method includes generating a super resolution version of a single input video sequence in at least the temporal dimension by exploiting data recurrence within the input video sequence or with respect to an external database of example video sequences. A signal may be an image, a video sequence, an audio signal, etc.

Abstract: Devices, systems, and methods of blind deblurring and blind super-resolution utilizing internal patch recurrence. Small signal patches tend to repeat “as is” across multiple scales of a natural signal. This fractal-like behavior is utilized for signal processing tasks, including “Blind Deblurring” or “Blind Super-Resolution”, namely, removing signal blur or increasing signal resolution without a-priori knowledge of the underlying blur kernel. While the cross-scale patch recurrence is strong in signals taken under ideal conditions, the cross-scale patch recurrence significantly diminishes when the acquisition blur deviates from an ideal blur. These deviations from ideal patch recurrences are used for recovering the underlying (unknown) blur kernel. The correct blur kernel is recovered by seeking the kernel which maximizes the patch similarity across scales of a related “reference” signal.

Abstract: A method for measuring bi-directional similarity between a first signal of a first size and a second signal of a second size includes matching at least some patches of the first signal with patches of the second signal for data completeness, matching at least some patches of the second signal with patches of the first signal for data coherence, calculating the bi-directional similarity measure as a function of the matched patches for coherence and the matched patches for completeness and indicating the similarity between the first signal and the second signal. Another method generates a second signal from a first signal where the second signal is different than the first signal by at least one parameter. The method includes attempting to maximize a bi-directional similarity measure between the second signal and the first signal.

Abstract: Device, system, and method of visual inference by collaborative composition. A method of searching for similar target content-items for a plurality of source content-items, the method implementable on a computing device, comprises: for each of said plurality of source content-items, searching for similar target content-items in a dataset, while taking into account collaborative feedback exchanged among at least some of currently-searched said plurality of source content-items.

Abstract: Method and apparatus for inferring irregularities in query data relative to referential data includes attempting to compose the query data, like a puzzle, from large chunks of the referential data, and inferring irregularities in the query data based on at least the size of the matching chunks. The larger the size of a matching chunk, the more likely it is that its corresponding region in the query data is valid and not irregular. Regions in the query data which cannot be composed from the referential data, or can only be composed using small fragmented pieces and not large chunks of the referential data, are considered irregular. The method and apparatus is applicable to all types of signals, including images, video data, medical data, one-dimensional signals and multi-dimensional signals, and can be used to identify, inter alia, suspicious behaviors, suspicious objects, irregular patterns and defects in goods.

Abstract: A method of image processing is disclosed, the method implementable on an electronic device, the method comprising: calculating for an image patch within an image at least one patch-dependent content information; based on said at least one patch-dependent content information, determining a patch-dependent search region; searching said patch-dependent search region for one or more image patches that are similar to said image patch; and processing said image patch based on said similar image patches found in said patch-dependent search region.

Abstract: A method includes measuring the likelihood that two different space-time video segments could have resulted from a similar underlying motion field without computing the field. The method may be employed to identify locations in a video sequence where at least one behavioral phenomenon similar to that demonstrated in the video segment occurs. For example, the phenomenon might be a dynamic behavior, an action, a rigid motion, and/or a non-rigid motion.

Abstract: A method includes finding regions of a reference signal which provide at least one of: local evidence scores and a global evidence score. The local evidence scores indicate local similarity of the regions of the reference signal to regions of a query signal and the global evidence score defines the extent of a global similarity of the query signal to the reference signal. A media exploring device is also included which includes an importance encoder and a media explorer. The importance encoder generates importance scores of at least portions of digital media as a function of at least one of local evidence scores and global evidence scores. The media explorer enables exploring through the digital media according to (i) the importance scores, (ii) data associations/links induced by the evidence scores between different portions of the digital media. The device may also include a media player to play the digital media with adaptive speeds as a function of the importance scores.

Abstract: A method implementable on a computing device includes exploiting data redundancy to combine high frequency information from at least two different scales of an input signal to generate a super resolution version of said input signal. An alternative method includes exploiting recurrence of data from an input signal in at least two different scales of at least one reference signal to extract and to combine high frequency information from a plurality of scales of said at least one reference signal to generate a super resolution version of said input signal. An alternative method includes generating a super resolution version of a single input video sequence in at least the temporal dimension by exploiting data recurrence within the input video sequence or with respect to an external database of example video sequences. A signal may be an image, a video sequence, an audio signal, etc.

Abstract: A method includes finding regions of a reference signal which provide at least one of: local evidence scores and a global evidence score. The local evidence scores indicate local similarity of the regions of the reference signal to regions of a query signal and the global evidence score defines the extent of a global similarity of the query signal to the reference signal. A media exploring device is also included which includes an importance encoder and a media explorer. The importance encoder generates importance scores of at least portions of digital media as a function of at least one of local evidence scores and global evidence scores. The media explorer enables exploring through the digital media according to (i) the importance scores, (ii) data associations/links induced by the evidence scores between different portions of the digital media. The device may also include a media player to play the digital media with adaptive speeds as a function of the importance scores.

Abstract: A method for measuring bi-directional similarity between a first signal of a first size and a second signal of a second size includes matching at least some patches of the first signal with patches of the second signal for data completeness, matching at least some patches of the second signal with patches of the first signal for data coherence, calculating the bi-directional similarity measure as a function of the matched patches for coherence and the matched patches for completeness and indicating the similarity between the first signal and the second signal. Another method generates a second signal from a first signal where the second signal is different than the first signal by at least one parameter. The method includes attempting to maximize a bi-directional similarity measure between the second signal and the first signal.

Abstract: A method includes matching at least portions of first, second signals using local self-similarity descriptors of the signals. The matching includes computing a local self-similarity descriptor for each one of at least a portion of points in the first signal, forming a query ensemble of the descriptors for the first signal and seeking an ensemble of descriptors of the second signal which matches the query ensemble of descriptors. This matching can be used for image categorization, object classification, object recognition, image segmentation, image alignment, video categorization, action recognition, action classification, video segmentation, video alignment, signal alignment, multi-sensor signal alignment, multi-sensor signal matching, optical character recognition, image and video synthesis, correspondence estimation, signal registration and change detection. It may also be used to synthesize a new signal with elements similar to those of a guiding signal synthesized from portions of the reference signal.

Abstract: A method and system for computing at least one of spatial and temporal relationships between at least first and second sequences of representations having respective first and second temporal progressions and including employing the first and second temporal progressions to obtain at least one of the spatial and temporal relationships.

Abstract: A method and system for aligning in at least one of time and space temporally ordered sequences of images comprising receiving a plurality of sequences of images, each sequence including a multiplicity of images, each plurality of sequences defining a space-time volume, and providing an output indication relating at least one point in a space-time volume corresponding to one of the plurality of sequences to at least one point in a space-time volume corresponding to at least another one of the plurality of sequences.

Abstract: A system and method for increasing space or time resolution of an image sequence by combination of a plurality input sources with different space-time resolutions such that the single output displays increased accuracy and clarity without the calculation of motion vectors. This system of enhancement may be used on any optical recording device, including but not limited to digital video, analog video, still pictures of any format and so forth. The present invention includes support for example for such features as single frame resolution increase, combination of a number of still pictures, the option to calibrate spatial or temporal enhancement or any combination thereof, increased video resolution by using high-resolution still cameras as enhancement additions and may optionally be implemented using a camera synchronization method.

Abstract: Method and apparatus for inferring irregularities in query data relative to referential data includes attempting to compose the query data, like a puzzle, from large chunks of the referential data, and inferring irregularities in the query data based on at least the size of the matching chunks. The larger the size of a matching chunk, the more likely it is that its corresponding region in the query data is valid and not irregular. Regions in the query data which cannot be composed from the referential data, or can only be composed using small fragmented pieces and not large chunks of the referential data, are considered irregular. The method and apparatus is applicable to all types of signals, including images, video data, medical data, one-dimensional signals and multi-dimensional signals, and can be used to identify, inter alia, suspicious behaviors, suspicious objects, irregular patterns and defects in goods.