Abstract: A method of determining delamination in a transistor is disclosed including loading a grey scale image of an transistor into memory, generating a black and white image based on the loaded grey scale image, identifying boundaries within the generated black and white image, cropping the black and white image based on the identified boundaries, identifying at least one feature in the cropped black and white image based on the identified boundaries, normalizing the cropped black and white image based on an attribute of the identified at least one feature, cropping the grey scale image based on the normalized black and white image, comparing the cropped grey scale image to a baseline grey scale image of the transistor, and determining a change in a percentage of delamination of the transistor between the baseline grey scale image and the cropped grey scale image based on the comparison.

Abstract: An endoscope system captures images of biological tissue when illuminated with first to third light, and generates first to third color image data. The amount of hemoglobin in the biological tissue and the oxygen saturation of the hemoglobin are calculated using components of the first to third color image data, an oxygen saturation distribution image that shows the distribution of the oxygen saturation is generated, and control is performed such that the generated oxygen saturation distribution image is displayed superimposed on a captured image of the biological tissue on a display. The oxygen saturation is calculated based on the amount of hemoglobin and a second ratio obtained using components of the color image data. The transparency is adjusted for pixels in which the value of the second ratio is outside an allowable range for the second ratio determined according to the amount of hemoglobin.

Abstract: A high resolution optical image is acquired by irradiating a mask with light emitted by a light source via a high resolution optical system. A low resolution optical image is acquired by irradiating the same mask with the light via a low resolution optical system. The design data of the mask pattern is corrected in light of shapes and dimensions determined according to at least one of a manufacturing process of the mask and a manufacturing process of a semiconductor device to be manufactured by transferring the mask pattern to a semiconductor wafer. Reference image data are generated corresponding to the high resolution optical image and the low resolution optical image. Whether the defect detected in the high resolution optical image is true or false is determined according to information of the defect detected in the low resolution optical image.

Abstract: Systems and methods for performing OCR of a series of images depicting text symbols.

Abstract: A method and device for processing and accessing image are provided. The image processing method includes: one or more protected areas are determined in an original picture to be processed, image data of the one or more protected areas are protected by adding access rights, one or more independent sub-image data corresponding to the one or more protected areas are obtained, and then the one or more protected areas in the original picture are shielded to obtain main image data, the images in the protected areas being invisible in the shielded original picture; and the one or more obtained independent sub-image data and the obtained main image data are associatively stored.

Abstract: An image processing apparatus, comprises an image acquiring unit that acquires a plurality of object images captured by imaging an object for a plurality of times; a shape acquiring unit that acquires shape information, which is information representing a shape of the object; an image dividing unit that divides each of the plurality of object images into at least two partial images, based on the shape information; and an image registering unit that registers the divided partial images.

Abstract: A method for facial recognition encode analysis comprises providing a training set of Gabor encoded arrays of face images from a database; and, for each encode array in the training set, evaluating the Gabor data to determine the accuracy of the fiducial points on which the encode array is based. The method also comprises training an outlier detection algorithm based on the evaluation of the encode arrays to obtain a decision function for a strength of accuracy of fiducial points in the encode arrays; and outputting the decision function for application to an encode array to be tested.

Abstract: An approach is provided for a redundant feature detection engine. The approach, for instance, involves segmenting an input image into a plurality of grid cells for processing by the redundant feature detection engine. The redundant feature detection engine includes a neural network. The approach also involves, for each of the plurality of grid cells, initiating a prediction of an object code by the redundant feature detection engine. The object code is a predicted feature that uniquely identifies an object depicted in the input image. The approach further involves aggregating the plurality of grid cells into one or more clusters based on the object code predicted for said each grid cell. The approach further involves predicting one or more features of the object corresponding to a respective cluster of the one or more clusters by merging one or more feature prediction outputs of said each grid cell in the respective cluster.

Abstract: An image similarity determination apparatus includes a memory and a processor configured to acquire a first image and a second image, perform selection of a first group and a second group from a plurality of feature points included in the first image and perform selection of a third group and a fourth group from a plurality of feature points included in the second image, calculate feature quantity for each feature point included in the first group and the third group on the basis of luminance and calculate feature quantity for each feature point included in the second group and the fourth group on the basis of hue, and determine similarity between the first image and the second image on the basis of both first comparison of the first group with the third group and second comparison of the second group with the fourth group.

Abstract: Systems and methods for improving video understanding tasks based on higher-order object interactions (HOIs) between object features are provided. A plurality of frames of a video are obtained. A coarse-grained feature representation is generated by generating an image feature for each of for each of a plurality of timesteps respectively corresponding to each of the frames and performing attention based on the image features. A fine-grained feature representation is generated by generating an object feature for each of the plurality of timesteps and generating the HOIs between the object features. The coarse-grained and the fine-grained feature representations are concatenated to generate a concatenated feature representation.

Abstract: Provided is a method of processing image data and detecting a region of an image represented by the image data to be excluded from an analysis of the image. According to the method, image data captured by a medical modality is received. An evaluation of a portion of the image data representing a two-dimensional view of a subject appearing in the image is conducted to locate, in the two-dimensional view, the region to be excluded from the analysis of the image. A feature pertinent to the analysis appearing in a remaining portion of the image, that is outside of the region to be excluded from the analysis located by the evaluation, is detected.

Abstract: A device that determines whether a first person extracted from a first image is identified with a second person extracted from a second image, includes a processor configured to extract a first feature value of a contour portion of the first person, from the first image, extract a second feature value of a contour portion of the second person, from the second image, and determine whether the first person is identified with the second person, based on comparison processing in which a first effect on a first comparison result of the first feature value and the second feature value becomes greater than a second effect on a second comparison result of a third feature value of a portion further from a contour than the contour portion of the first person and a fourth feature value of a portion further from the contour than the contour portion of the second person.

Abstract: Apparatuses and methods for image processing are provided. The image processing apparatus performs area classification and object detection in an image, and includes a feature map generator configured to generate the feature map of the input image using the neural network, and an image processor configured to classify the areas and to detect the objects in the image using the generated feature map.

Abstract: A system and method for automated contrast arrival detection in temporally phased images or datasets of tissues effectively determines contrast arrival in regions that are substantially free of arteries. A plurality of tissue voxels in a plurality of temporally phased images are identified as a function of voxel enhancement characteristics associated with discrete tissue voxels. A processor/process computes average enhancement characteristics from the plurality of identified tissue voxels. The average enhancement characteristics are compared with predetermined average enhancement characteristics associated with contrast media arrival phases. Contrast media arrival phases in the temporally phased images are provided based on the comparison.

Abstract: An image processing apparatus includes an identification unit, a calculator, and an image processor. The identification unit identifies a group of pixels indicating an object of interest in medical image data. The calculator specifies a cross-section corresponding to the running direction of the object of interest indicated by the group of pixels in the medical image data. The image processor performs noise reduction on the cross-section identified.

Abstract: A neural network recognition method includes obtaining a first neural network that includes layers and a second neural network that includes a layer connected to the first neural network, actuating a processor to compute a first feature map from input data based on a layer of the first neural network, compute a second feature map from the input data based on the layer connected to the first neural network in the second neural network, and generate a recognition result based on the first neural network from an intermediate feature map computed by applying an element-wise operation to the first feature map and the second feature map.

Abstract: An authentication method and corresponding apparatus includes obtaining iris images, and constituting an enroll set including iris codes and mask codes corresponding to the iris images. The authentication method and corresponding apparatus also include generating a synthesized code including a synthesized iris code and a synthesized mask code based on correlations between the iris codes included in the enroll set in block units.

Abstract: Provided are an apparatus and method for extracting a salient line for an information sign. The apparatus includes an image input portion configured to receive an image including information indication lines to be detected, a signal calculator configured to generate a symmetric function signal consistent with a thickness and a direction of an information indication line having a preset thickness in the received image and perform a convolution of the information indication line and the generated symmetric function signal, a highest value enhancement portion configured to enhance highest values in results of the convolution using a signal quality improvement filter, and a local highest value extractor configured to find local highest values among the enhanced highest values and extract a line connecting the local highest values as a salient line connecting center points in the information indication line.

Abstract: This disclosure relates to improved techniques for performing computer vision functions on medical images, including object segmentation functions for identifying medical objects in the medical images and grading functions for determining severity labels for medical conditions exhibited in the medical images. The techniques described herein utilize a neural network architecture to perform these and other functions. The neural network architecture can be trained, at least in part, using semi-supervised learning techniques that enable the neural network architecture to accurately perform the object segmentation and grading functions despite limited availability of pixel-level annotation information.

Abstract: An apparatus for identifying a candidate area in a first image corresponding to an object in a second image, includes a memory and a processor to divide the plurality of candidate areas into a plurality of small candidate areas, divide an image area of the object into a plurality of small areas, perform first comparison processing for a first part, when there is a first candidate area lacking image information of the small candidate area corresponding to the first part, perform second comparison processing for a second part, predict missing result on the small candidate area corresponding to the first part in the first candidate area based on a result of the first comparison processing on a candidate area other than the first candidate area, and a result of the second comparison processing on the plurality of candidate areas, and identify the candidate area based on a prediction.