This application discloses methods, apparatus, and electronic devices for face pose estimation and three-dimensional face reconstruction. The face pose estimation method comprises: acquiring a two-dimensional face image for processing, constructing a three-dimensional face model corresponding to the two-dimensional face image, and determining a face pose of the two-dimensional face image based on face feature points of the three-dimensional face model and face feature points of the two-dimensional face image. With this approach, the face pose estimation is performed based on the three-dimensional face model corresponding to the two-dimensional face image, instead of only based on a three-dimensional average face model. As a result, a high accuracy pose estimation can be obtained even for a face with large-angle and exaggerated facial expressions. Thus, robustness of the pose estimation can be effectively improved.

The present disclosure relates to an image processing method. The method may include: obtaining image data; reconstructing an image based on the image data, the image including one or more first edges; obtaining a model, the model including one or more second edges corresponding to the one or more first edges; matching the model and the image; and adjusting the one or more second edges of the model based on the one or more first edges.

A method for image segmentation may include acquiring an image including a region of interest (ROI). The ROI has a first margin, the ROI includes a subregion, and the subregion has a second margin. The method may further include acquiring a first model according to the ROI, wherein the first model has a third margin. The method may further determine, based on the first margin and the third margin, a second model by matching the first model with the image, wherein the second model includes a sub-model, and the sub-model has a fourth margin. The method may further include determining, based on the second margin, a third model by adjusting the fourth margin of the sub-model in the second model. The method may further include segmenting the ROI according to the third model and generating a segmented ROI based on a result of the segmentation.

Automated and objective methods for quantifying a retinal characteristic from a retinal Optometric Coherence Tomography retinal image, and methods for detecting occult ocular pathology, diagnosing ocular pathology, reducing age-bias in OCT image analysis, and monitoring efficacy ocular/retinal disease therapies based on the quantification method are disclosed.

Disclosed is a method for modelling age-related traits of a face, from a picture of the face, wherein the age-related traits are either wrinkles or age spots, the method including: for each age-related trait of the face of the same nature, generating a vector including parameters of shape and appearance of the trait; and generating, from the generated vectors, a single representation vector modeling the age-related traits of the same nature in the face. The single representation vector stores information regarding the number of traits in the face and joint probabilities, over the face, of the shape and appearance features of the traits.

Embodiments of the present invention provide techniques for coding video data efficiently based on detection of objects within video sequences. A video coder may perform object detection on the frame and when an object is detected, develop statistics of an area of the frame in which the object is located. The video coder may compare pixels adjacent to the object location to the object's statistics and may define an object region to include pixel blocks corresponding to the object's location and pixel blocks corresponding to adjacent pixels having similar statistics as the detected object. The coder may code the video frame according to a block-based compression algorithm wherein pixel blocks of the object region are coded according to coding parameters generating relatively high quality coding and pixel blocks outside the object region are coded according to coding parameters generating relatively lower quality coding.

Techniques and systems are provided for maintaining blob trackers for one or more video frames. For example, a blob tracker is identified for a current video frame. The blob tracker is associated with a blob detected for the current video frame. The blob includes pixels of at least a portion of one or more objects in the current video frame. A current characteristic of pixels in a region of the current video frame associated with the blob tracker is determined. A previous characteristic of pixels in a region of a previous video frame associated with the blob tracker is also determined. A difference is determined between the current characteristic and the previous characteristic, and a status of the blob tracker is determined based on the determined difference. The status of the blob tracker indicating whether to maintain the blob tracker for the one or more video frames.

A method for image segmentation may include acquiring an image including a region of interest (ROI). The ROI has a first margin, the ROI includes a subregion, and the subregion has a second margin. The method may further include acquiring a first model according to the ROI, wherein the first model has a third margin. The method may further determine, based on the first margin and the third margin, a second model by matching the first model with the image, wherein the second model includes a sub-model, and the sub-model has a fourth margin. The method may further include determining, based on the second margin, a third model by adjusting the fourth margin of the sub-model in the second model. The method may further include segmenting the ROI according to the third model and generating a segmented ROI based on a result of the segmentation.

Intelligent image parsing for anatomical landmarks and/or organs detection and/or segmentation is provided. A state space of an artificial agent is specified for discrete portions of a test image. A set of actions is determined, each specifying a possible change in a parametric space with respect to the test image. A reward system is established based on applying each action of the set of actions and based on at least one target state. The artificial agent learns an optimal action-value function approximator specifying the behavior of the artificial agent to maximize a cumulative future reward value of the reward system. The behavior of the artificial agent is a sequence of actions moving the agent towards at least one target state. The learned artificial agent is applied on a test image to automatically parse image content.

A method and an electronic device for object tracking in a sequence of light-field captures. A data acquisition unit acquires a sequence of light-field captures, wherein each light-field capture comprises a plurality of views. A feature determining unit determines features of an initial visual appearance model for an object of interest in a reference view of a first light-field capture. A feature matching unit matches the features in the reference view and in the further views of the first light-field capture. A feature discarding unit discards features that cannot be well matched in all views of the first light-field capture. An appearance model building unit builds an updated visual appearance model for the object of interest based on the remaining features. Finally, a movement tracking unit tracks the movement of the object of interest in the sequence of light-field captures using the visual appearance model.