Abstract: Techniques for multistage process model training are described herein. Another aspect includes determining a first gray level histogram corresponding to a first input image. Another aspect includes determining a second gray level histogram corresponding to a second input image. Another aspect includes determining a set of change values, each change value corresponding to a change in a respective gray level from the first gray level histogram to the second gray level histogram. Another aspect includes comparing each change value of the set of change values to a threshold. Another aspect includes, based on determining that a first change value of the set of change values is higher than the threshold, adding a first gray level corresponding to the first change value to a hot zone of the second input image. Another aspect includes training a model using the hot zone of the second input image.

Abstract: Provided is a method of reconstructing an integral image based on computational integral imaging reconstruction (CIIR), the method including determining a crop ratio of an original elemental image array (EIA), cropping each elemental image in the original EIA according to the determined crop ratio to generate a cropped and translated EIA, magnifying each elemental image in the cropped and translated EIA according to an adjustable magnification factor (k), overlapping the magnified elemental images with an overlapping number that is adjusted according to the adjustable magnification factor (k), and normalizing the overlapping elemental images to generate a reconstructed image.

Abstract: Embodiments of this application provide an image processing method performed at a computing device. The image processing method includes: obtaining a to-be-processed image with ghost reflection; calculating an image gradient of the to-be-processed image; determining, according to the image gradient, a gradient of a target image obtained after the ghost reflection is removed from the to-be-processed image; and generating the target image based on the gradient of the target image. According to the technical solution of the embodiments of this application, the ghost reflection in the image can be effectively removed, ensuring high quality of a processed image.

Abstract: A face image processing method and apparatus, an electronic device, and a storage medium are provided. The method includes: obtaining face key points and a face deflection angle in a face image; determining a submalar triangle center in the face image according to the face key points and the face deflection angle; determining a submalar triangle region in the face image according to the face key points and the submalar triangle center; and performing color filling on the submalar triangle region. By means of the present disclosure, a submalar triangle region can be accurately positioned, and submalar triangle filling is performed based on the accurately positioned submalar triangle region, thereby obtaining a more natural filling effect.

Abstract: Interpretation maps of deep neural networks are provided that use Renyi differential privacy to guarantee the robustness of the interpretation. In one aspect, a method for generating interpretation maps with guaranteed robustness includes: perturbing an original digital image by adding Gaussian noise to the original digital image to obtain m noisy images; providing the m noisy images as input to a deep neural network; interpreting output from the deep neural network to obtain m noisy interpretations corresponding to the m noisy images; thresholding the m noisy interpretations to obtain a top-k of the m noisy interpretations; and averaging the top-k of the m noisy interpretations to produce an interpretation map with certifiable robustness.

Abstract: In one example in accordance with the present disclosure, a system is described. The system includes a pose estimator to identify a plurality of anatomical points on a person depicted in an image. A frame former of the system generates a frame for the person by connecting a first set of the plurality of anatomical points to form a skeleton and forming at least a head region and a torso region of the person based on a second set of the plurality of anatomical points. A map generator of the system generates a spatial gradient map projecting outwards from the frame. In the system, the spatial gradient map is based on pixel distance from the frame and an intensity along the gradient map indicates a pixels likelihood of form a part of the person.

Abstract: Chrominance data channels are reconstructed from a color-mosaic image data input and de-noised with guidance from a luminance data channel extracted from that same mosaic data input to produce de-noised chrominance channels. The de-noised chrominance channels are combined with a de-noised version of the luminance channel to yield a primary-color-channel output image.

Abstract: Forensic method for identifying forged documents. For each of a stream of incoming jpeg images, using a processor configured for determining whether jpeg image/s is a replacement forgery by determining whether a first portion of individual image which resides at a known location (known likely to be replaced by forger) within the individual jpeg image has been replaced, including: indicator, face-djpg, for the first portion at known location; computing indicator, aka nonface-djpg, for a second portion of individual image which resides at a comparison location within the jpeg image known as unlikely to be replaced by a forger; and determining whether face-djpg and nonface-djpg fulfill predetermined logical criterion and deciding whether the individual jpeg image is a replacement forgery accordingly.

Abstract: Provided are methods and apparatuses for positioning face feature points. The method includes: carrying out edge detection on a face image to obtain a face feature line image; and fusing the face image and the face feature line image to obtain position information of face feature points.

Abstract: An image processing apparatus according to the present invention includes at least one memory and at least one processor Which function as: a setting unit configured to he capable of setting any of a plurality of processing modes which include a first processing mode to display an image having a first brightness range and a second processing mode to display an image having a second brightness range which is narrower than the first brightness range and a processing unit configured to generate frame image data, which is a display target, in a first state in which image processing in accordance with a currently set processing mode is performed, and generates a capture image data corresponding to the frame image data in a second state after changing the first state in accordance with the currently set processing mode.

Abstract: A method and system for face alignment. The method may include obtaining an image processing model set including M (M?2) candidate models, and obtaining a test image including a target face. The method may also include conducting T (T?1) stages of model set updating operation. Each stage of the T stages of model set updating operation may include conducting a performance evaluation to each candidate model of the image processing model set with respect to the test image, and updating the image processing model set by excluding at least one model from the image processing model set based on the performance evaluation. The method may further include designating, after completing the T stages of model set updating operation, at least one candidate model of the image processing model set as a target model, and determining, based on the target model, a result shape as a shape of the target face.

Abstract: Image processing methods and apparatus are described. The image processing method comprises receiving input of a visible-ray image and an infrared-ray image obtained by photographing a same subject, estimating, based on the visible-ray image, the infrared-ray image and motion information, a blur estimate associated with the visible-ray image, and generating, based on the estimated blur estimate, a corrected visible-ray image.

Abstract: An automated method and apparatus are provided for identifying when a first video is a content-identical variant of a second video. The first and second video each include a plurality of image frames, and the image frames of either the first video or the second video include at least one black border. A plurality of variants are generated of selected image frames of the first video and the second video. The variants are then compared to each other, and the first video is identified as being a variant of the second video when at least one match is detected among the variants.

Abstract: Systems, methods, and computer-readable media are provided for foreground image segmentation. In some examples, a method can include obtaining a first image of a target and a second image of the target, the first image having a first field-of-view (FOV) and the second image having a second FOV; determining, based on the first image, a first segmentation map that identifies a first estimated foreground region in the first image; determining, based on the second image, a second segmentation map that identifies a second estimated foreground region in the second image; generating a third segmentation map based on the first segmentation map and the second segmentation map; and generating, using the second segmentation map and the third segmentation map, a refined segmentation mask that identifies the target as a foreground region of the first and/or second image.

Abstract: An image processing method for a computer device. The method includes obtaining a to-be-processed image belonging to a first image category; inputting the to-be-processed image into a first stage image conversion model, to obtain a first intermediate image; and converting the first intermediate image into a second intermediate image through a second stage image conversion model. The method also includes determining a first weight matrix corresponding to the first intermediate image; determining a second weight matrix corresponding to the second intermediate image; and fusing the first intermediate image and the second intermediate image according to the corresponding first weight matrix and second weight matrix, to obtain a target image corresponding to the to-be-processed image and belonging to a second image category. A sum of the first weight matrix and the second weight matrix being a preset matrix.

Abstract: An image processing apparatus comprises: at least one processor or circuit configured to perform the operations of following units: an obtaining unit configured to obtain a plurality of image signals being composed of image signals shot wider different shooting conditions; a determination unit configured to determine a composition rate in order to composite an image signal that has the predetermined shooting condition with a noise-reduced image that has the predetermined shooting condition, in accordance with an inter-frame change amount and a parameter indicating the predetermined shooting condition; a noise reduction unit configured to composite the image signal with the noise-reduced image using the composition rate to generate a new noise-reduced image; and a composition unit configured to composite the new noise-reduced image and an image signal that has another shooting condition.

Abstract: An apparatus for ascertaining, from at least one image, a delimited region and/or a motion path of at least one object includes at least one artificial neural network (ANN) made up of several successive layers. The first layer of the ANN receives as input the at least one image or a part thereof. The second layer supplies as output a boundary line of the delimited region, a linear course of the motion path, or a portion of the boundary line or motion path. The dimensionality of the second layer is lower than the dimensionality of the first layer.

Abstract: A facial feature matching system comprises a facial feature matching engine. A first user selection of reference facial images is received, and facial features of reference faces are characterized using the facial feature recognition engine comprising a neural network with input, hidden, and output layers. The facial features are weighted. The weighted facial features are used to identify users that have facial features similar to the weighted facial features, wherein the respective reference faces include faces different than the faces of the users. Similarity indicators are generated for the identified users. The generated respective similarity indicators are used to generate a ordering of the identified users which is rendered via the user device. A first user selection of a second user in the ordered identified users is received and the first user and the second user are enabled to communicate over an electronic communication channel.

Abstract: This disclosure relates to techniques for the robust usage of semantic segmentation information in image processing techniques, e.g., shallow depth of field (SDOF) renderings. Semantic segmentation may be defined as a process of creating a mask over an image, wherein pixels are segmented into a predefined set of semantic classes. Segmentations may be binary (e.g., a ‘person pixel’ or a ‘non-person pixel’) or multi-class (e.g., a pixel may be labelled as: ‘person,’ ‘dog,’ ‘cat,’ etc.). As semantic segmentation techniques grow in accuracy and adoption, it is becoming increasingly important to develop methods of utilizing such segmentations and developing flexible techniques for integrating segmentation information into existing computer vision applications, such as synthetic SDOF renderings, to yield improved results in a wide range of image capture scenarios.

Abstract: When one looks at a face, one cannot help but ‘read’ it: in the blink of an eye, people form reliable impressions of both transient psychological states (e.g., happiness) and stable character traits (e.g., trustworthiness). Such impressions are irresistible, formed with high levels of consensus, and important for social decisions. Disclosed herein is a large-scale data-driven methodology that allows for the easy manipulation of social trait information in hyper-realistic face images. For example, a given face image could be made to look more or less trustworthy by moving a simple slider. Further, this method can not only generate faces, but can ‘read’ faces as well, providing confidence estimates of different social traits for any arbitrary image. The disclosed approach is both fast and accurate, and represents a paradigm shift in facial photo manipulation.