Abstract: Examples are described for applying different settings for image capture to different portions of image data. For example, an image sensor can capture image data of a scene and can send the image data to an image signal processor (ISP) and a classification engine for processing. The classification engine can determine that a first object image region depicts a first category of object, and a second object image region depicts a second category of object. Different confidence regions of the image data can identify different degrees of confidence in the classifications. The ISP can generate an image by applying a different settings to the different portions of the image data. The different portions of the image data can be identified based on the object image regions and confidence regions.

Abstract: Examples are described for applying different settings for image capture to different portions of image data. For example, an image sensor can capture image data of a scene and can send the image data to an image signal processor (ISP) and a classification engine for processing. The classification engine can determine that a first object image region depicts a first category of object, and a second object image region depicts a second category of object. Different confidence regions of the image data can identify different degrees of confidence in the classifications. The ISP can generate an image by applying a different settings to the different portions of the image data. The different portions of the image data can be identified based on the object image regions and confidence regions.

Abstract: The present disclosure generally relates to image processing. For example, aspects of the present disclosure include systems and techniques for performing spatial and temporal processing of image data. Certain aspects provide an apparatus for processing frame data. The apparatus generally includes a memory, and one or more processors coupled to the memory, the one or more processors configured to: perform a first noise reduction operation based on first frame data via a machine learning component to generate first processed frame data; generate first feedback data based on the first processed frame data; and perform, via the machine learning component, a second noise reduction operation based on second frame data and the first feedback data.

Abstract: Techniques and systems are provided for processing one or more frames or images. For instance, a process for determining exposure for one or more frames includes obtaining a motion map for one or more frames. The process includes determining, based on the motion map, motion associated with the one or more frames of a scene. The motion corresponds to movement of one or more objects in the scene relative to a camera used to capture the one or more frames. The process includes determining, based on the determined motion, a number of frames and an exposure for capturing the number of frames. The process further includes sending a request to capture the number of frames using the determined exposure duration.

Abstract: Techniques and systems are provided for machine-learning based image stabilization. In some examples, a system obtains a sequence of frames captured by an image capture device during a period of time, and collects motion sensor measurements calculated by a motion sensor associated with the image capture device based on movement of the image capture device during the period of time. The system generates, using a deep learning network and the motion sensor measurements, parameters for counteracting motions in one or more frames in the sequence of frames, the motions resulting from the movement of the image capture device during the period of time. The system then adjusts the one or more frames in the sequence of frames according to the parameters to generate one or more adjusted frames having a reduction in at least some of the motions in the one or more frames.

Abstract: Examples are described for applying different settings for image capture to different portions of image data. For example, an image sensor can capture image data of a scene and can send the image data to an image signal processor (ISP) and a classification engine for processing. The classification engine can determine that a first object image region depicts a first category of object, and a second object image region depicts a second category of object. Different confidence regions of the image data can identify different degrees of confidence in the classifications. The ISP can generate an image by applying a different settings to the different portions of the image data. The different portions of the image data can be identified based on the object image regions and confidence regions.

Abstract: Techniques and systems are provided for machine-learning based image stabilization. In some examples, a system obtains a sequence of frames captured by an image capture device during a period of time, and collects motion sensor measurements calculated by a motion sensor associated with the image capture device based on movement of the image capture device during the period of time. The system generates, using a deep learning network and the motion sensor measurements, parameters for counteracting motions in one or more frames in the sequence of frames, the motions resulting from the movement of the image capture device during the period of time. The system then adjusts the one or more frames in the sequence of frames according to the parameters to generate one or more adjusted frames having a reduction in at least some of the motions in the one or more frames.

Abstract: Techniques and systems are provided for machine-learning based image stabilization. In some examples, a system obtains a sequence of frames captured by an image capture device during a period of time, and collects motion sensor measurements calculated by a motion sensor associated with the image capture device based on movement of the image capture device during the period of time. The system generates, using a deep learning network and the motion sensor measurements, parameters for counteracting motions in one or more frames in the sequence of frames, the motions resulting from the movement of the image capture device during the period of time. The system then adjusts the one or more frames in the sequence of frames according to the parameters to generate one or more adjusted frames having a reduction in at least some of the motions in the one or more frames.

Abstract: Techniques and systems are provided for machine-learning based image stabilization. In some examples, a system obtains a sequence of frames captured by an image capture device during a period of time, and collects motion sensor measurements calculated by a motion sensor associated with the image capture device based on movement of the image capture device during the period of time. The system generates, using a deep learning network and the motion sensor measurements, parameters for counteracting motions in one or more frames in the sequence of frames, the motions resulting from the movement of the image capture device during the period of time. The system then adjusts the one or more frames in the sequence of frames according to the parameters to generate one or more adjusted frames having a reduction in at least some of the motions in the one or more frames.

Abstract: A method, apparatus, and manufacture for encoding a video sequence is provided. Encoding the video sequence includes performing a weighted prediction estimation between a reference frame of the video sequence and a target frame of the video sequence. Performing the weighted prediction includes performing an initial weighted prediction estimation for each block of the reference frame. Next, blocks are clustered according to their initial weighted prediction estimates. Then, each block of the target image is assigned to a corresponding region based on the clustering. During the video encoding, weighted-prediction is employed for each block according to its corresponding region.

Abstract: A system, method, and computer program product for digital stabilization of video data from cameras producing multiple simultaneous views, typically from rolling shutter type sensors, and without requiring a motion sensor. A first embodiment performs an estimation of the global transformation on a single view and uses this transformation for correcting other views. A second embodiment selects a distance at which a maximal number of scene points is located and considers only the motion vectors from these image areas for the global transformation. The global transformation estimate is improved by averaging images from several views and reducing stabilization when image conditions may cause incorrect stabilization. Intentional motion is identified confidently in multiple views. Local object distortion may be corrected using depth information. A third embodiment analyzes the depth of the scene and uses the depth information to perform stabilization for each of multiple depth layers separately.

Abstract: A system, method, and computer program product for saturation insensitive weighted prediction coefficients estimation for video coding. The saturated pixels in frames of an input video are excluded from consideration in estimating weighted prediction coefficients that are then used to encode the input video. The saturated pixels include those having luma and/or chroma values above or below limits, and may be identified by use of a histogram. The saturated pixels may also be found by analyzing a lower resolution version of the frames, and/or sampling pixels in the frames. Embodiments enable improved video quality for a given bit rate, and increased coding efficiency, particularly for input video including frames with rapid pixel value changes. Separate portions of the frames may be processed separately, and the embodiments may execute within an optimization loop that minimizes a predetermined error function. The encoding format may include the H.264/AVC standard.

Abstract: Embodiments may be directed to lens cameras which may be cameras arranged as a sensor in a lens cap. A lens camera may comprise a printed circuit board with a digital image sensor and associated components enclosed in a cylindrical body that may be constructed of metal, plastic, or the like, or combination thereof. Lens cameras may be fitted with lens mounts for attaching host devices, cameras, interchangeable lens, or the like. Lens mounts on a lens camera may be arranged to be compatible with one or more standard lens mounts. Accordingly, a lens camera may be attached to cameras that have compatible lens mounts. Also, interchangeable lens having lens mounts compatible with the lens camera may be attached to the lens camera. Further, lens cameras may communicate with host devices using wired or wireless communication facilities.

Abstract: Embodiments may be directed to lens cameras which may be cameras arranged as a sensor in a lens cap. A lens camera may comprise a printed circuit board with a digital image sensor and associated components enclosed in a cylindrical body that may be constructed of metal, plastic, or the like, or combination thereof. Lens cameras may be fitted with lens mounts for attaching host devices, cameras, interchangeable lens, or the like. Lens mounts on a lens camera may be arranged to be compatible with one or more standard lens mounts. Accordingly, a lens camera may be attached to cameras that have compatible lens mounts. Also, interchangeable lens having lens mounts compatible with the lens camera may be attached to the lens camera. Further, lens cameras may communicate with host devices using wired or wireless communication facilities.

Abstract: A method, apparatus, and manufacture for encoding a video sequence is provided. During a first exploitation phase of an encoding pass of the video encoding, macro-blocks are encoded employing at least one encoding parameter. At least one encoder statistic is evaluated based on the encoding of macro-blocks during the first exploitation phase. Next, during a first exploration phase of the encoding pass, macro-blocks are encoded employing at least one encoding parameter that is different than the encoding parameter(s) used during the first exploitation phase. At least one encoder statistic is evaluated based on the encoding of the macro-blocks during the first exploitation phase. The encoder statistic(s) based on the first exploration stage is compared when the encoder statistic(s) based on the first exploitation phase, and a steady-state value of the encoding parameter(s) is updated based on the comparison. The process may then continue to alternate between exploration and exploitation phases.

Abstract: A method and device are provided for method for stabilization of image data by an imaging device. In one embodiment, a method includes detecting image data for a first frame and a second frame, performing motion estimation to determine one or more motion vectors associated with global frame motion for image data of the first frame, performing an outlier rejection function to select at least one of the one or more motion vectors, and determining a global transformation for image data of the first frame based, at least in part, on motion vectors selected by the outlier rejection function. The method may further include determining a stabilization transformation for image data of the first frame by refining the global transformation to correct for unintentional motion and applying the stabilization transformation to image data of the first frame to stabilize the image data of the first frame.

Abstract: In a digital camera or other image acquisition device, motion vectors between successive image frames of an object scene are calculated from normalized values of pixel luminance in order to reduce or eliminate any effects on the motion calculation that might occur when the object scene is illuminated from a time varying source such as a fluorescent lamp. Calculated motion vectors are checked for accuracy by a robustness matrix.

Abstract: A method and device are provided for method for stabilization of image data by an imaging device. In one embodiment, a method includes detecting image data for a first frame and a second frame, performing motion estimation to determine one or more motion vectors associated with global frame motion for image data of the first frame, performing an outlier rejection function to select at least one of the one or more motion vectors, and determining a global transformation for image data of the first frame based, at least in part, on motion vectors selected by the outlier rejection function. The method may further include determining a stabilization transformation for image data of the first frame by refining the global transformation to correct for unintentional motion and applying the stabilization transformation to image data of the first frame to stabilize the image data of the first frame.

Abstract: In a digital camera or other image acquisition device, motion vectors between successive image frames of an object scene are calculated from normalized values of pixel luminance in order to reduce or eliminate any effects on the motion calculation that might occur when the object scene is illuminated from a time varying source such as a fluorescent lamp. Calculated motion vectors are checked for accuracy by a robustness matrix.