Abstract: A plurality of training iterations is conducted for a machine learning task. A given iteration includes generating a version of a stacking model using a portion of a labeled data set. Proposed labels are then obtained in the iteration using the generated version of the stacking model for a set of unlabeled records. The unlabeled records and their proposed labels are then used to generate versions of base models for the iteration. After the training iterations are completed, a trained ensemble of models including the stacking model and the base models is stored.

Abstract: Query types for which responses are to be generated with respect to records comprising text attributes are identified, including a text interpretation query type for which records may comprise one or more response-contributor strings. Results of the text interpretation query for a record are based at least partly on an extracted-property class of the records. A machine learning model comprising a first sub-model and a second sub-model is trained to extract results of the text interpretation query. The first sub-model generates an extracted-property class for a record, and the second sub-model predicts positions of response-contributor strings within the record based at least in part on the extracted-property class. A trained version of the model is stored.

Abstract: Similarity detection methods and systems are provided that utilize a convolutional neural network model to jointly learn string matching and semantic textual similarity as an image recognition solution. For example, in some embodiments described herein, the similarity detection system may receive two strings as input, transform the two strings into two separate vectors, generate a high-resolution image and a low-resolution image, apply one or more convolutional operations to each image, and determine string matching and semantic textual similarity based at least partly on the high-resolution image and the low-resolution image.

Abstract: A method includes projecting motion vectors describing a transformation from a previous video frame to a future video frame onto a plane between the previous video frame and the future video frame, detecting potential artifacts at the plane based on an intersection of a cover region and an uncover region on the plane, and analyzing a dissimilarity between a trial video frame and both the previous video frame and the future video frame. The trial video frame is generated between the previous video frame and the future video frame based on a frame rate conversion ratio derived from a source frame rate and a desired frame rate. The method also includes estimating reliability of the projected motion vectors based on the potential artifact detection and the dissimilarity analysis.

Abstract: A system, method, and computer program product are provided for refining motion vectors. In operation, a plurality of motion vectors associated with a current frame and a first resolution are created. Furthermore, the motion vectors are refined utilizing information including at least one of first information describing motion vectors associated with a previous frame and second information describing motion vectors associated with the current frame and a second resolution.

Abstract: A method and apparatus is provided for adjusting the contrast of an input image. A blurred image corresponding to the input image is obtained, the brightness level of at least some of the pixels in the input image being varied to provide the blurred image. A ring likelihood for pixels in the input image is obtained, the ring likelihood providing a measure of the likelihood that said pixels are non-edge pixels in the neighborhood of an edge pixel. The output image is produced as a sum of the brightness levels in the input image and the blurred image in dependence on the ring likelihoods.

Abstract: A method and apparatus adjusts the contrast of an input image formed of pixels. Each pixel in the input image has an input brightness level. At least some of the pixels in the output or enhanced image have an output brightness level that is different from their input brightness level. Brightness level data of the input image is filtered to obtain a high frequency component of the input image in a single operational stage. The high frequency component is amplified. The original image and the amplified high frequency component are summed to produce brightness level data of the output image.

Abstract: A method includes projecting motion vectors describing a transformation from a previous video frame to a future video frame onto a plane between the previous video frame and the future video frame, detecting potential artifacts at the plane based on an intersection of a cover region and an uncover region on the plane, and analyzing a dissimilarity between a trial video frame and both the previous video frame and the future video frame. The trial video frame is generated between the previous video frame and the future video frame based on a frame rate conversion ratio derived from a source frame rate and a desired frame rate. The method also includes estimating reliability of the projected motion vectors based on the potential artifact detection and the dissimilarity analysis.

Abstract: A composite video signal has a chrominance signal (14) and a luminance signal (10). A method and apparatus are disclosed for adjusting the chrominance signal (14) to obtain an enhanced chrominance signal (15) in which at least some transitions in the enhanced chrominance signal (15) align more closely with edges in the luminance signal (10) than transitions in the original chrominance signal (14) align with edges in the luminance signal (10). An edge in the luminance signal (10) of the composite video signal is detected (11). The position of the edge relative to a current pixel is detected (11). The chrominance of the current pixel is adjusted (13) in accordance with at least the position of said current pixel relative to the edge to obtain an enhanced chrominance signal (15).

Abstract: A method and apparatus for adjusting the contrast of an image, in which each pixel in the image has an initial brightness level that is within a range of brightness levels between a minimum brightness level and a maximum brightness level. For each brightness level within the range, the number of pixels that have the same initial brightness level is counted. The pixels are divided into at least two types by identifying the or each set of pixels having initial brightness levels in which each of a plurality of adjacent ones of said brightness levels has more pixels than a threshold number. The pixels of the plurality of adjacent ones of said brightness levels are treated as a first type. The remaining pixels are treated as a second type. For the or each set of pixels of the first type, contrast enhancement is carried out separately on those pixels.

Abstract: A method and apparatus is disclosed for obtaining interpolation filters for use in a method or apparatus for enhancing the spatial resolution of a digital image. In the method or apparatus for enhancing the spatial resolution, input pixels from a low resolution image are classified and interpolated using the interpolation filters to form a high resolution image. The method for obtaining the interpolation filters includes obtaining a low resolution image from a high resolution source image. A plurality of pixels in the low resolution image are classified with respect to interpolation filters so that each of the plurality of pixels in the low resolution image is allocated to a respective interpolation filter. The interpolation filters are updated. The classifying and updating steps are repeated until a convergence criterion is met.

Abstract: A method and apparatus for adjusting the contrast of an input image is disclosed. Brightness level data of the input image is filtered to obtain at least a low frequency component and a high frequency component. The high frequency component is amplified. The input image and the amplified high frequency component are summed to produce brightness level data of the output image. The filtering is carried out by at least one recursive infinite impulse response filter having a single delay coefficient that is adaptive to areas of high contrast in the input image.

Abstract: A method and apparatus is provided for adjusting the contrast of an input image. A blurred image corresponding to the input image is obtained, the brightness level of at least some of the pixels in the input image being varied to provide the blurred image. A ring likelihood for pixels in the input image is obtained, the ring likelihood providing a measure of the likelihood that said pixels are non-edge pixels in the neighbourhood of an edge pixel. The output image is produced as a sum of the brightness levels in the input image and the blurred image in dependence on the ring likelihoods.

Abstract: A method and apparatus adjusts the contrast of an input image formed of pixels. Each pixel in the input image has an input brightness level. At least some of the pixels in the output or enhanced image have an output brightness level that is different from their input brightness level. Brightness level data of the input image is filtered to obtain a high frequency component of the input image in a single operational stage. The high frequency component is amplified. The original image and the amplified high frequency component are summed to produce brightness level data of the output image.

Abstract: A composite video signal has a chrominance signal (14) and a luminance signal (10). A method and apparatus are disclosed for adjusting the chrominance signal (14) to obtain an enhanced chrominance signal (15) in which at least some transitions in the enhanced chrominance signal (15) align more closely with edges in the luminance signal (10) than transitions in the original chrominance signal (14) align with edges in the luminance signal (10). An edge in the luminance signal (10) of the composite video signal is detected (11). The position of the edge relative to a current pixel is detected (11). The chrominance of the current pixel is adjusted (13) in accordance with at least the position of said current pixel relative to the edge to obtain an enhanced chrominance signal (15).

Abstract: A method and apparatus for adjusting the contrast of an image, in which each pixel in the image has an initial brightness level that is within a range of brightness levels between a minimum brightness level and a maximum brightness level. For each brightness level within the range, the number of pixels that have the same initial brightness level is counted. The pixels are divided into at least two types by identifying the or each set of pixels having initial brightness levels in which each of a plurality of adjacent ones of said brightness levels has more pixels than a threshold number. The pixels of the plurality of adjacent ones of said brightness levels are treated as a first type. The remaining pixels are treated as a second type. For the or each set of pixels of the first type, contrast enhancement is carried out separately on those pixels.

Abstract: A method and apparatus for adjusting the contrast of an input image is disclosed. Brightness level data of the input image is filtered to obtain at least a low frequency component and a high frequency component. The high frequency component is amplified. The low frequency component and the amplified high frequency component are summed to produce brightness level data of the output image. The filtering is carried out by at least one recursive infinite impulse response filter having a single delay coefficient that is adaptive to areas of high contrast in the input image.

Abstract: A method and apparatus is disclosed for obtaining interpolation filters for use in a method or apparatus for enhancing the spatial resolution of a digital image. In the method or apparatus for enhancing the spatial resolution, input pixels from a low resolution image are classified and interpolated using the interpolation filters to form a high resolution image. The method for obtaining the interpolation filters includes obtaining a low resolution image from a high resolution source image. A plurality of pixels in the low resolution image are classified with respect to interpolation filters so that each of the plurality of pixels in the low resolution image is allocated to a respective interpolation filter. The interpolation filters are updated. The classifying and updating steps are repeated until a convergence criterion is met.