Abstract: This specification describes a method comprising generating, from a plurality of first images representing a scene, at least one stereoscopic panoramic image comprising a left-eye panoramic image and a right-eye panoramic image. Depth map images are generated corresponding to each of the left and right-eye panoramic images. Each of the left and right-eye panoramic images are re-projected to obtain a plurality of second images, each associated with a respective virtual camera. Each of the left and right-eye depth map images are re-projected to generate a re-projected depth map associated with each second image. A first three-dimensional model of the scene based on the plurality of second images is determined. A second three-dimensional model of the scene based on the plurality of re-projected depth map images is determined. One or more corresponding points of the first and second three-dimensional models is or are compared to determine a scaling factor.

Abstract: A method, an apparatus and a computer program product are provided, wherein the method comprises receiving a video comprising video frames as an input; generating set of object proposals from the video, the set of object proposals comprising positive object proposals and negative object proposals; generating object tracklets comprising regions appearing in consecutive frames of the video, said regions corresponding to object proposals with a high confidence; constructing a graph for the object proposals to rescore the object proposals in the generated object tracklets; and aggregating the rescored object proposals to produce an object detection.

Abstract: A method comprising performing image re-projection on each of a plurality of first images of a scene, thereby to generate a plurality of re-projected second images of the scene, wherein each first image of the scene is captured by a respective camera of a first multi-directional image capture apparatus and each second image of the scene is associated with a respective virtual camera; processing the plurality of second images based on a previously generated virtual three dimensional model of the scene, thereby to generate respective positions of the virtual cameras associated with the second images; and determining a position of the first multi-directional image capture apparatus based on one or more of the generated positions of the virtual cameras.

Abstract: The invention relates to a method, an apparatus and a computer program product for analyzing media content. The method comprises receiving media content objects by a feature extractor for extracting a plurality of feature maps from said media content objects; processing the plurality of feature maps in a bidirectional Long-Short Term memory neural network, where the bidirectional Long-Short Term memory neural network is aligned along different directions of the feature maps to produce low resolution feature maps; upsampling the low resolution feature maps to the size of received media content; and assigning each pixel of the upsampled feature maps with a label of maximum likelihood for segmenting objects from the upsampled feature maps.

Abstract: The invention relates to a method and an apparatus implementing the method. The method comprises extracting region proposals from a media content; selecting a set of region proposals corresponding to an object in the media content; identifying objects of interest; determining an object-specific representation by an iterative tracking method; sampling positive examples from the set of tracked region proposal groups obtained from the iterative tracking method; and performing object segmentation.

Abstract: In an example embodiment a method, apparatus and computer program product are provided. The method includes extracting a first set of target-object regions and at least one set of non-target object regions from a plurality of regions of a video content based at least on likelihood information. A plurality of unlabeled regions of the video content are classified based on the first set of target-object regions and the at least one set of non-target object regions to generate a second set of target object regions that is denser than the first set of target object regions. A model is learned for modelling at least one of the target object and non-target objects of the video content based at least on the second set of target object regions. The target object in the video content is segmented based on the model and the second set of target object regions.

Abstract: There is provided methods, apparatuses and computer program products for image segmentation in which pixel data of an image and information of a first set of pixels of the image indicative of pixels belonging to a foreground of the image are obtained. A color model of the image is estimated. To obtain a geodesic distance map at least one geodesic distance from a pixel of the first set of pixel to another pixel of the image which is not in the first set of pixels is determined. An energy function is optimized on the basis of the geodesic distance map to obtain a segmentation of the image.

Abstract: In an example embodiment, a method, apparatus and computer program product are provided. The method includes facilitating receipt of a plurality of superpixels of an input image and an object selection input for selecting part of a foreground object in the input image. The method includes determining first set of superpixels as first set of foreground seeds and second set of superpixels as first set of background seeds. The method includes filtering the first set of background seeds to generate second set of background seeds based on geodesic distances. The method includes adding superpixels of first set of unclassified superpixels to the second set of background seeds to generate third set of background seeds, and adding superpixels of second set of unclassified superpixels to the third set of background seeds to generate fourth set of background seeds based on applying a classifier on the second set of unclassified superpixels.

Abstract: The invention relates to a method, comprising: receiving a plurality of images, wherein the plurality of images comprises content that relates to a same object; preprocessing said more than one images to form a feature vector for each region in an image; discovering object-like regions from each image by means of the feature vectors; determining an object appearance model for each image according to the object-like regions; generating an object hypotheses by means of the object appearance model; segmenting the same object in the plurality of images to generate segmented objects; and generating a multiple view segmentation according to the segmented objects.

Abstract: In an example embodiment a method, apparatus and computer program product are provided. The method includes extracting a first set of target-object regions and at least one set of non-target object regions from a plurality of regions of a video content based at least on likelihood information. A plurality of unlabeled regions of the video content are classified based on the first set of target-object regions and the at least one set of non-target object regions to generate a second set of target object regions that is denser than the first set of target object regions. A model is learned for modelling at least one of the target object and non-target objects of the video content based at least on the second set of target object regions. The target object in the video content is segmented based on the model and the second set of target object regions.

Abstract: In an example embodiment, a method, apparatus and computer program product are provided. The method includes facilitating receipt of a plurality of superpixels of an input image and an object selection input for selecting part of a foreground object in the input image. The method includes determining first set of superpixels as first set of foreground seeds and second set of superpixels as first set of background seeds. The method includes filtering the first set of background seeds to generate second set of background seeds based on geodesic distances. The method includes adding superpixels of first set of unclassified superpixels to the second set of background seeds to generate third set of background seeds, and adding superpixels of second set of unclassified superpixels to the third set of background seeds to generate fourth set of background seeds based on applying a classifier on the second set of unclassified superpixels.

Abstract: There is provided methods, apparatuses and computer program products for image segmentation in which pixel data of an image and information of a first set of pixels of the image indicative of pixels belonging to a foreground of the image are obtained. A colour model of the image is estimated. To obtain a geodesic distance map at least one geodesic distance from a pixel of the first set of pixel to another pixel of the image which is not in the first set of pixels is determined. An energy function is optimized on the basis of the geodesic distance map to obtain a segmentation of the image.