Abstract: According to an embodiment of the disclosure, a method performed by an apparatus may include obtaining a plurality of images, each of the plurality of images comprising a view of a scene. The method may include computing respective scores for each of the plurality of images. The method may include estimating respective camera poses of each of the plurality of images. The method may include using the computed scores and the estimated camera poses to determine a new camera pose useable for generating a new image comprising a view of the scene and having a score greater than a first threshold score. The method may include generating the new image using the new camera pose.

Abstract: Broadly speaking, embodiments of the present techniques provide a method for personalising a trained speaker verification machine learning, ML, model for a specific user, on-device (i.e. on the end user device which is going to be used to run the personalised ML model). Advantageously, the present techniques improve the personalisation of the ML model on-device without requiring large volumes of data to be stored on the device.

Abstract: A computer-implemented image manipulation apparatus and method (200) configured to receiving an input image (202) and a desired style. The method can obtain a representation (204) of the input image selected from a plurality of stored representations of a plurality of images, wherein each said representation comprises data describing a set of image features. The method can modify image features in the obtained representation to correspond to the input image and/or the desired style to produce a modified representation (207), and render a reference image (209) based on the modified representation. A manipulated image is generated by performing a style transfer operation (210) on the input image using the rendered reference image. Embodiments may access a data store to find a group of stored images based on similarity between image content descriptors of groups of stored images and those of an input image to retrieve a stored reference image.

Abstract: Broadly speaking, the present techniques generally relates to methods, systems and apparatuses for performing image orientation correction, i.e. correcting or adjusting images that are tilted. In particular, the present application relates to a computer-implemented method for generating a training dataset for training a machine learning, ML, model using federated learning to perform image orientation correction, and methods for training the ML model using the generated training dataset. Advantageously, the method to generate a training dataset enables a diverse training dataset to be generated while maintaining user data privacy, where the diversity refers to the range of image tilt angles represented by the dataset. The present application also provides methods for training the ML model using the generated training dataset.

Abstract: A computer-implemented federated learning method is disclosed. The method comprises: for each of a number, n, of clients: determining a diversity score of a dataset corresponding to that client for training a machine learning model, wherein the diversity score is a measure of dataset variability; aggregating, weighted by the respective diversity score, models corresponding to each of the clients; and sending the aggregated model to at least one receiving client.

Abstract: Broadly speaking, embodiments of the present techniques provide methods for performing shadow detection and/or removal in images. In particular, the present techniques provide a computer-implemented method for generating a synthetic training dataset for training a machine learning, ML, model using federated learning to perform shadow detection (and optionally removal), and methods for training the ML model using the generated training dataset. Advantageously, the method to generate a training dataset enables a diverse training dataset to be generated while maintaining user data privacy, where the diversity refers to the variety of scenes containing shadows.

Abstract: Image processing comprises receiving content from an application; rendering the content into a rendered frame and storing the rendered frame in a frame buffer; transferring the rendered frame from the frame buffer to a compositor, and displaying the rendered frame using the compositor. The image processing further comprises applying a non-linear transformation to at least part of the rendered frame during the rendering the content into the rendered frame; during the transferring of the rendered frame to the compositor, or during the displaying of the rendered frame by the compositor.

Abstract: Embodiments disclosed herein relate to device and method for processing frames. For example, a buffer of a device is arranged to store a plurality of rendered frames rendered at a frame rendering rate and a time stamp for each of rendered frames. A compositor of a device is arranged to obtain a timestamp of a synchronisation signal for synchronising the display of frames with a display refresh rate. In response to obtaining a timestamp of a synchronisation signal, a compositor is arranged to trigger access to a buffer to obtain two rendered frames having timestamps closest to a timestamp of a synchronisation signal. An interpolator of a device is arranged to generate an interpolated rendered frame for display by performing an interpolation operation using two rendered frames. An interpolation operation takes into account the difference between timestamps of each of two rendered frames and a timestamp of a synchronisation signal.

Abstract: Broadly speaking, this disclosure generally relates to methods, systems and apparatuses for performing monocular depth estimation, i.e. depth estimation using a single camera. In particular, this disclosure relates to a method for generating a training dataset for training a machine learning, ML, model using federated learning to perform depth estimation. Advantageously, the method to generate a training dataset enables a diverse training dataset to be generated while maintaining user data privacy. This disclosure also provides methods for training the ML model using the generated training dataset. Advantageously, the methods determine whether a community ML model that is trained by client devices needs to be retrained, and/or whether a global ML model, which is used to generate the community ML model, needs to be retrained.

Abstract: A method and system for processing an image is described. For example, the method comprises detecting a plurality of objects within an in input image; identifying dimensions of a display on which the input image is to be displayed; cropping the input image to obtain a cropped image which matches the identified dimensions, wherein the cropped image includes at least one of the plurality of detected objects; obtaining a list of missing objects which are not visible in the cropped image and which were detected in the input image; outputting a representation of each missing object in the list of missing objects to be displayed together with the cropped image; generating an updated image comprising the representation of at least one missing object and which matches the identified dimensions; and outputting the updated image to be displayed on the display.

Abstract: Apparatus for determining configuration settings for a display apparatus includes a processor and a memory storing instructions executable by the processor where the processor is configured to determine configuration settings for a display apparatus in dependence on one or more first input parameters indicative of one or more conditions that affect a user experience when a user views content reproduced on the display apparatus, predict an expected user response as a result of operating the display apparatus according to the determined configuration settings, search for an updated set of configuration settings that are expected to produce a user response closer to an intended result than the predicted user response, and control the display apparatus to operate according to the updated set of configuration settings.

Abstract: An image processing method is disclosed, comprising comparing a first image and a modified second image, the modified second image comprising a copy of the first image in which at least one pixel value of one or more pixels have been modified, and based on the result of the comparison, generating first image reconstruction information which can be combined with the modified second image to reconstruct the first image. A corresponding image reconstruction method comprises combining the image reconstruction information with the modified second image to reconstruct the first image. Apparatus for implementing the methods is also disclosed.

Abstract: A computer-implemented image manipulation apparatus and method (200) configured to receiving an input image (202) and a desired style. The method can obtain a representation (204) of the input image selected from a plurality of stored representations of a plurality of images, wherein each said representation comprises data describing a set of image features. The method can modify image features in the obtained representation to correspond to the input image and/or the desired style to produce a modified representation (207), and render a reference image (209) based on the modified representation. A manipulated image is generated by performing a style transfer operation (210) on the input image using the rendered reference image. Embodiments may access a data store to find a group of stored images based on similarity between image content descriptors of groups of stored images and those of an input image to retrieve a stored reference image.

Abstract: Provided is an electronic device for stabilizing a 360-degree video, the electronic device including a memory storing one or more instructions, and a processor for executing the one or more instructions stored in the memory, wherein the processor is configured to execute the one or more instructions to: when a 360-degree video is reproduced, allocate probability values to a plurality of pixels included in a frame of the 360-degree video, based on a possibility that each of the plurality of pixels is included in a user's field of view (FOV), determine a three-dimensional (3D) rotation for the 360-degree video, based on the allocated probability values, and generate a stabilized 360-degree video by applying the 3D rotation to the 360-degree video.

Abstract: Embodiments disclosed herein relate to device and method for processing frames. For example, a buffer of a device is arranged to store a plurality of rendered frames rendered at a frame rendering rate and a time stamp for each of rendered frames. A compositor of a device is arranged to obtain a timestamp of a synchronisation signal for synchronising the display of frames with a display refresh rate. In response to obtaining a timestamp of a synchronisation signal, a compositor is arranged to trigger access to a buffer to obtain two rendered frames having timestamps closest to a timestamp of a synchronisation signal. An interpolator of a device is arranged to generate an interpolated rendered frame for display by performing an interpolation operation using two rendered frames. An interpolation operation takes into account the difference between timestamps of each of two rendered frames and a timestamp of a synchronisation signal.

Abstract: Provided is a method of processing a 360-degree image that includes: obtaining a plurality of motion vectors for a 360-degree image, determining at least one motion vector, which indicates global rotation of the 360-degree image, among the plurality of motion vectors through filtering; performing three-dimensional-transformation on the determined at least one motion vector so as to acquire three-dimensional rotation information of the 360-degree image, and correcting distortion, of the 360-degree image, caused by shaking, based on the acquired three-dimensional rotation information.

Abstract: Apparatus for determining configuration settings for a display apparatus includes a processor and a memory storing instructions executable by the processor where the processor is configured to determine configuration settings for a display apparatus in dependence on one or more first input parameters indicative of one or more conditions that affect a user experience when a user views content reproduced on the display apparatus, predict an expected user response as a result of operating the display apparatus according to the determined configuration settings, search for an updated set of configuration settings that are expected to produce a user response closer to an intended result than the predicted user response, and control the display apparatus to operate according to the updated set of configuration settings.

Abstract: An image processing method is disclosed, comprising comparing a first image and a modified second image, the modified second image comprising a copy of the first image in which at least one pixel value of one or more pixels have been modified, and based on the result of the comparison, generating first image reconstruction information which can be combined with the modified second image to reconstruct the first image. A corresponding image reconstruction method comprises combining the image reconstruction information with the modified second image to reconstruct the first image. Apparatus for implementing the methods is also disclosed.

Abstract: Apparatus and methods are provided. The method includes receiving content that comprises at least one 360 degree image. The method also includes applying a rotation to the at least one 360 degree image to increase a compression ratio, when a compression algorithm compresses the at least one 360 degree image that is rotated, relative to another compression ratio achieved when the compression algorithm compresses the content without the applied rotation to the at least one 360 degree image. The method also includes compressing the content including the at least one 360 degree image that is rotated based on the compression algorithm. The method further includes outputting the compressed content and a rotation parameter that defines the applied rotation.

Abstract: Provided is an electronic device for stabilizing a 360-degree video, the electronic device including a memory storing one or more instructions, and a processor for executing the one or more instructions stored in the memory, wherein the processor is configured to execute the one or more instructions to: when a 360-degree video is reproduced, allocate probability values to a plurality of pixels included in a frame of the 360-degree video, based on a possibility that each of the plurality of pixels is included in a user's field of view (FOV), determine a three-dimensional (3D) rotation for the 360-degree video, based on the allocated probability values, and generate a stabilized 360-degree video by applying the 3D rotation to the 360-degree video.