Techniques are provided in which a plurality of edges are detected within a digital image. An anchor point located along an edge of the plurality of edges is selected. An analysis grid associated with the anchor point is generated, the analysis grid including a plurality of cells. An anchor point normal vector comprising a normal vector of the edge at the anchor point is calculated. Edge pixel normal vectors comprising normal vectors of the edge at locations along the edge within the cells of the analysis grid are calculated. A histogram of similarity is generated for each of one or more cells of the analysis grid, each histogram of similarity being based on a similarity measure between each of the edge pixel normal vectors within a cell and the anchor point normal vector, and a descriptor is generated for the analysis grid based on the histograms of similarity.

Techniques for interactively determining/visualizing the color content of a source image and how the corresponding image data is mapped to device colors are described herein. For example, the color content of a digital image can be converted between different color spaces to identify gamut limitations of an output device (e.g., a printing assembly), discover color(s) that cannot be accurately reproduced, etc. Color space conversions enable the transformation of the color content of the digital image from device-specific colorants to a device-independent representation (and vice versa). In some embodiments, these transformations are facilitated using lookup tables that are implemented in graphical processing unit-resident memory.

This facial authentication device is provided with: a detecting means for detecting a plurality of facial feature point candidates, using a plurality of different techniques, for at least one facial feature point of a target face, from a plurality of facial images containing the target face; a reliability calculating means for calculating a reliability of each facial image, from statistical information obtained on the basis of the plurality of detected facial feature point candidates; and a selecting means for selecting a facial image to be used for authentication of the target face, from among the plurality of facial images, on the basis of the calculated reliabilities.

Techniques are provided for multi-modal sensitive recognition. A digital data set for an object is obtained according to a modality, where the digital data set includes digital representations of the object at different values of a dimension of relevance of the modality. A reference location associated with the object is identified. A modal descriptor is derived for the modality according to an implementation of a multi-modal recognition algorithm by deriving a set of feature descriptors for the reference location and at the different values of the corresponding dimension of relevance, calculating a set of differences between the feature descriptors in the set of feature descriptors, and aggregating the set of differences into the modal descriptor. A device is then configured to initiate an action as a function of the modal descriptor.

A subject tracking device includes a first histogram generation unit configured to generate a first histogram representing an appearance frequency of an image feature quantity of a subject region, for each segment and for each auxiliary segment of the image feature quantity, a second histogram generation unit configured to generate, for each candidate region, a second histogram representing an appearance frequency of the image feature quantity of an image of a candidate region for each segment and for each auxiliary segment of the image feature quantity, a difference derivation unit configured to compare the first histogram with the second histogram for each segment and for each auxiliary segment, and derive, for each candidate region, a total difference value of the appearance frequencies of the image feature quantities of the first histogram and the second histogram; and a region selection unit configured to select the candidate region having a minimum total difference value from among the candidate regions. The auxiliary segment is defined to straddle boundaries between the segments.

According to an embodiment of the present disclosure, there is provided a method of predicting a brain tissue lesion distribution, the method including: a model learning operation of learning a prediction model for predicting a brain tissue lesion distribution of a subject by using brain image data of a plurality of previous patients; an input obtaining operation of obtaining input data from brain image data of the subject; and an output operation of generating output image data including information on the lesion distribution after recanalization treatment for the subject, by using the prediction model. The prediction model includes a success prediction model that is learned by using data of patients in which recanalization treatment is successful among the plurality of previous patients, and a failure prediction model that is learned by using data of patients in which recanalization treatment fails among the plurality of previous patients.

The present disclosure provides a method and apparatus for detecting a region of interest in a video, a device and a storage medium. The method may include: acquiring a current to-be-processed frame from a picture frame sequence of a video; detecting a region of interest (ROI) in the current to-be-processed frame, in response to determining that the current to-be-processed frame is a detection picture frame, to determine at least one ROI in the current to-be-processed frame; and updating a to-be-tracked ROI, based on the ROI in the current to-be-processed frame and a tracking result determined by a pre-order tracking picture frame; and tracking the current to-be-processed frame based on the existing to-be-tracked ROI, in response to determining that the current to-be-processed frame is a tracking picture frame, to determine at least one tracking result as the ROI of the current to-be-processed frame.

Systems and methods are provided for image classification using histograms of oriented gradients (HoG) in conjunction with a trainer. The efficiency of the process is greatly increased by first establishing a bitmap which identifies a subset of the pixels in the HoG window as including relevant foreground information, and limiting the HoG calculation and comparison process to only the pixels included in the bitmap.

A high-definition map system receives sensor data from vehicles travelling along routes and combines the data to generate a high definition map for use in driving vehicles, for example, for guiding autonomous vehicles. A pose graph is built from the collected data, each pose representing location and orientation of a vehicle. The pose graph is optimized to minimize constraints between poses. Points associated with surface are assigned a confidence measure determined using a measure of hardness/softness of the surface. A machine-learning-based result filter detects bad alignment results and prevents them from being entered in the subsequent global pose optimization. The alignment framework is parallelizable for execution using a parallel/distributed architecture. Alignment hot spots are detected for further verification and improvement. The system supports incremental updates, thereby allowing refinements of sub-graphs for incrementally improving the high-definition map for keeping it up to date

A method for detecting a human skeleton is provided. The method includes: receiving a video frame, wherein the video frame comprises a human body; determining whether the video frame comprises prediction information; determining whether a first intra-coded macroblock (IMB) ratio of a target area comprising the human body in the video frame is greater than a first threshold when the video frame comprises the prediction information; and using a motion vector (MV) to estimate skeleton information of the human body when the first IMB ratio of the target area is not greater than the first threshold.