Abstract: A machine learning system receives a reference image and generates a series of projected view images of a physical object represented in the images. Parallel neural networks may receive the reference image and series of projected view images for analysis to determine one or more features of the physical object. By pooling the results from the parallel network, a single output may be provided to a set of decodes that are trained to identify a material property of the one or more items. As a result, a three-dimensional model may be generated that includes a graphical representation of the object as a function of its material properties to enable improved rendering.

Abstract: A motion estimation method and apparatus for video coding of a multi-view sequence is described. In one embodiment, a motion estimation method includes identifying one or more pixels in a first frame of a multi-view video sequence, and constraining a search range associated with a second frame of the multi-view video sequence based on an indication of a desired correlation between efficient coding and semantic accuracy. The semantic accuracy relies on use of geometric configurations of cameras capturing the multi-view video sequence. The method further includes searching the second frame within the constrained search range for a match of the pixels identified in the first frame.

Abstract: An image processing apparatus and an image processing method, applied to e.g. a TV conference system or a TV telephone system, in which an image to be transmitted or received is captured and reconstructed into a virtual view point image which may appear as if it is captured by a virtual camera. The correspondence between pixels may be taken with higher robustness and higher accuracy, while mismatch in such correspondence, otherwise caused due to disparity, may be reduced to as small a value as possible. To this end, the disparity between the images, obtained on imaging an object from different view points by at least two cameras 11a, 12a, are detected and, based on the detected disparity, each of the images is separated into plural image regions. The relation of correspondence is found of the separated each image region, in association with the object, from one pixel position on the same horizontal lines to the next.

Abstract: The invention indexes an audiovisual recording to enable navigation within the audiovisual recording. In particular, the invention provides indexing within a “show” of a recording of an audiovisual broadcast (e.g., a television broadcast or a radio/satellite audio broadcast) that enables navigation within the show, in contrast to indexing among shows (i.e., identification of locations corresponding to the start and/or end of shows) that can enable navigation between shows. The invention also enables navigation among one or more intelligently selected segments of the recording in a non-linear manner. A graphical user interface can be implemented to effect navigation in accordance with the invention of a recording of audiovisual content.

Abstract: An image processing apparatus and an image processing method, applied to e.g. a TV conference system or a TV telephone system, in which an image to be transmitted or received is captured and reconstructed into a virtual view point image which may appear as if it is captured by a virtual camera. The correspondence between pixels may be taken with higher robustness and higher accuracy, while mismatch in such correspondence, otherwise caused due to disparity, may be reduced to as small a value as possible. To this end, the disparity between the images, obtained on imaging an object from different view points by at least two cameras 11a, 12a, are detected and, based on the detected disparity, each of the images is separated into plural image regions. The relation of correspondence is found of the separated each image region, in association with the object, from one pixel position on the same horizontal lines to the next.

Abstract: A motion estimation method and apparatus for video coding of a multi-view sequence is described. In one embodiment, a motion estimation method includes identifying one or more pixels in a first frame of a multi-view video sequence, and constraining a search range associated with a second frame of the multi-view video sequence based on an indication of a desired correlation between efficient coding and semantic accuracy. The semantic accuracy relies on use of geometric configurations of cameras capturing the multi-view video sequence. The method further includes searching the second frame within the constrained search range for a match of the pixels identified in the first frame.