Abstract: Methods, device and data stream format are disclosed in the present document for the encoding, the formatting and the decoding of depth information representative of a 3D scene. Compression and decompression of quantized values by a video codec leads to a value error. This error on values is particularly sensitive for depth encoding. The present invention proposes to encode and decode depth with a quantization function that minimize an angle error when a value error on quantized depth creates a location delta between the projected point and the de-projected point. The inverse of such a quantization function has to be encoded in metadata associated with the 3D scene, for example as a LUT, to be retrieved at the decoding, as such functions are not tractable.

Abstract: This invention relates to a self-supervised privacy preservation action recognition system leveraging a learnable transformation anonymization function. The system is designed to process videos by removing spatial cues to protect privacy while retaining critical information for action recognition. The anonymization function, based on an encoder-decoder model, undergoes iterative training on a dataset to optimize the balance between obscuring privacy-sensitive information and preserving capability of recognizing actions. This training involves freezing and adjusting the weights of an action recognition branch and a self-supervised privacy removal branch to refine the model's effectiveness. The outcome is an anonymized video with minimized privacy information leakage, suitable for action analysis without privacy labels.

Abstract: A method of modulating a depth of a hologram, the method includes: obtaining hologram data; determining a scale factor based on a hardware specification of a holographic display to display a three-dimensional (3D) hologram image in a space by using the hologram data; and modulating depth information of the hologram data based on the scale factor.

Abstract: Example implementations include a method, apparatus and computer-readable medium for person detection by a computer device, comprising receiving a plurality of video frames from a camera located in an environment, wherein the video frames include a depiction of a person in the environment. The implementations further include identifying attributes of the person from the video frames and comparing the attributes with attribute entries in a database of suspicious persons, wherein the database of suspicious persons includes attributes of a plurality of persons associated with an alert in the environment. The implementations include retrieving a profile indicative of an alert history of the suspicious person in response to determining a match between the attributes of the person and an entry in the database corresponding to a suspicious person, and transmitting, to a second computer device, the retrieved profile and a notification that indicates that the suspicious person is in the environment.

Abstract: A method and apparatus for parallel context processing for example for high coding efficient entropy coding in HEVC. The method comprising retrieving syntax element relating to a block of an image, grouping at least two bins belonging to similar context based on the syntax element, and coding the grouped bins in parallel.

Abstract: A computing device performs a method of decoding video data by determining a co-located picture of the current coding unit; locating a spatial neighbor block of the current coding unit that corresponds to the co-located picture; determining a motion shift vector for the current coding unit from one or more motion vectors associated with the spatial neighbor block according to a predefined fixed order; and reconstructing a sub-block-based temporal motion vector for a respective sub-block of a plurality of sub-blocks in the current coding unit from a corresponding sub-block in the collocated picture based on the motion shift vector.

Abstract: A device may receive and store video frames in the memory. The device may define, in a video frame among the video frames, a plurality of bounding boxes to visually bound one or more subjects, each bounding box of the plurality of bounding boxes may be defined in accordance with a pre-defined aspect ratio, and each bounding box may comprise padding above and laterally with respect to the one or more subjects in the video frame. The device may reduce the plurality of bounding boxes by merging two bounding boxes among the plurality of bounding boxes to create a merged bounding box, wherein the merged bounding box may maintain the pre-defined aspect ratio and visually bounds the two bounding boxes being merged, and wherein the reduced plurality of bounding boxes is provided to an object detector.

Abstract: A system for detecting the presence of pests, the system comprising: at least one remote camera system, each of the at least one remote camera system comprising an image capture device coupled to a processor and a wireless transmitter, each of the at least one remote camera system having a pest detection surface and being configured to: capture one or more image of the pest detection surface; process the captured one or more image to recognise the potential presence of a target pest on the pest detection surface; and transmit data from the captured one or more image in response to recognising the potential presence of one or more target pests; and a server configured to: receive the transmitted data; process the transmitted data to verify the potential presence of the target pest on the pest detection surface; and provide an output indicating the verification.

Abstract: The monitoring method for wheat kernel moisture content in-field based on PlanetScope satellite imagery in this disclosure comprises: (1) data collection and database establishment; (2) feature engineering: constructing various broad-band indices on the original bands provided by PlanetScope to expand the feature domain, and evaluate the importance of each feature to rank the features based on the comprehensive importance calculated by permutation importance based on regressors; (3) model construction: progressively adding the spectral features one by one as independent variables to the GBDT model according to the ranking results, and screen out the optimal spectral feature domain and inversion model for wheat kernel moisture content; (4) wheat kernel moisture content mapping. The disclosure utilizes machine learning inversion models based on spectral features and PlanetScope satellite imagery to obtain the wheat kernel moisture content at any position of the target farmland.

Abstract: A device for efficient object detection and selective display of video frames is disclosed. A plurality of bounding boxes to visually bound one or more subjects are defined in the received video frames; each bounding box includes padding above and laterally with respect to the one or more subjects in the video frame. The device may reduce the plurality of bounding boxes by merging bounding boxes to create reduced plurality of merged bounding boxes. The device performs object detection on the reduced plurality of merged bounding boxes by searching a training database for detecting and identifying suspicious objects. The identified objects are annotated. On identifying that an object is suspicious, based on a trained database of firearms in a plurality of environments across a plurality of industries, the video frame is transmitted to a graphical user interface of a user device.

Abstract: When encoding/decoding a current block of a current picture using intra block copy (“BC”) prediction, the location of a reference block is constrained so that it can be entirely within an inner search area of the current picture or entirely within an outer search area of the current picture, but cannot overlap both the inner search area and the outer search area. In some hardware-based implementations, on-chip memory buffers sample values of the inner search area, and off-chip memory buffers sample values of the outer search area. By enforcing this constraint on the location of the reference block, an encoder/decoder can avoid memory access operations that are split between on-chip memory and off-chip memory when retrieving the sample values of the reference block. At the same time, a reference block close to the current block may be used for intra BC prediction, helping compression efficiency.

Abstract: A three-dimensional display device includes a display panel that displays a parallax image, an optical panel including a plurality of less-transmissive portions and a plurality of transmissive portions repeatedly arranged alternately in a parallax direction, and a controller that controls the display panel. Each of the plurality of transmissive portions has a first width in the parallax direction greater than a second width of each of the plurality of less-transmissive portions in the parallax direction. The controller causes the display panel to display a black image in a binocular viewable area on the display panel viewable to two eyes of a user to allow parallax separation of the parallax image.

Abstract: A method for video stream processing in a video chain is provided that includes transmitting a video stream in the video chain, receiving, by a first video node in the video chain, a region of interest (ROI) command from a second video node in the video chain, wherein the ROI command includes an ROI type indicator, and performing, by the first video node, the ROI command according to the ROI type indicator.

Abstract: A system and methods for ghost imaging second harmonic generation microscopy. Imaging data is collected in parallel, providing faster imagine reconstruction and enabling reconstruction in scattering environments. Ghost imaging, split light beam interacting with a target and a second light beam unimpeded and not required to pass through the same background. A second harmonic generation image is reconstructed from the detected photons.

Abstract: A process for identifying a road feature in an image includes receiving an image at a controller, identifying a region of interest within the image and converting the region of interest from red-green-blue (RGB) to a single color using the controller. A set edges is detected within the region of interest, and at least one line within the set of edges is identified using the controller. The at least one line is compared with a set of known and expected road marking features, and the set of at least one first line in the at least one line is identified as corresponding to a road feature in response to the at least the first line matching the set of known and expected road marking features.

Abstract: A method for coding implemented by a decoding device or an encoding device includes: obtaining a first prediction mode for a first sub-block, obtaining a second prediction mode for a second sub-block; obtaining a first prediction value for a sample in the current block according to the first prediction mode, obtaining a second prediction value for the sample in the current block according to the second prediction mode; obtaining a value of an indicator, the value of the indicator indicating a blending operation which is performed for the sample in the current block; and obtaining a combined prediction value for the sample in the current block, according to the value of the indictor, the first prediction value, and the second prediction value. The blending operation adaptively performs around a separation line dividing the current block into at least the first sub-block and the second sub-block.

Abstract: A video coding mechanism is disclosed. The mechanism includes receiving a bitstream comprising a current picture including a plurality of video coding layer (VCL) network abstraction layer (NAL) units that do not have a same NAL unit type. Active entries of reference picture lists are obtained for slices positioned in a sub-picture A (subpicA) in subsequent pictures following the current picture in decoding order. The active entries do not refer to any reference picture preceding the current picture in decoding order when the subpicA at the current picture is associated with an intra-random access point (IRAP) NAL unit type. The subsequent pictures are decoded based on the reference picture list active entries. The subsequent pictures are forwarded for display as part of a decoded video sequence.

Abstract: A method for detecting a position of a seatbelt in a vehicle includes: capturing, by a camera, a source image of an occupant in the vehicle; determining a pose of the occupant based on the source image; determining an occupant shadow based on the pose of the occupant; determining, based on the occupant shadow, a shadow overlying a region of interest (ROI) in the source image; and detecting, based on the shadow overlying the ROI, the seatbelt within the ROI. A system for detecting a position of a seatbelt is also provided. The system includes a camera and controller in communication with the camera and configured to: determine a pose of the occupant; determine an occupant shadow based on the pose of the occupant; determine, based on the occupant shadow, a shadow overlying the ROI; and detect, based on the shadow overlying the ROI, the seatbelt within the ROI.

Abstract: A video analysis technique is implemented using a modification of a Wasserstein Generative Adversarial Network (WGAN), named Spatial-Temporal Diverse Augmentation in Generative Models for Rare Event Synthesis (STAGES). A sequence of video images is received and features of the video images are identified. The STAGES technique is used to identify unusual or notable features found in frames of the video sequence as identified features, using spatial and temporal generators and discriminators to capture both spatial and temporal patterns in videos and thereby accurately detect rare events in complex situations. The video is then scanned for the identified features.

Abstract: A method for monitoring protected regions of a camera view includes receiving, at a processor of a video camera system, a video stream including a series of video frames depicting at least one scene. A background model is identified based on a predefined representation of at least one region of interest associated with the scene. Each of the video frames is compared to the background model, to define an associated binary mask. A persistent change in the at least one region of interest is detected, based on the binary masks, the persistent change having an associated duration that exceeds a predefined threshold. In response to detecting the persistent change, a change notification message including a video file is generated. The method also includes causing display of the change notification message via a user interface and/or causing transmission of the change notification message to a remote compute device.