A method for detecting security holograms on documents in a video stream is disclosed, including: searching for interest points and calculating descriptors in a frame; filtering of interest points in the previous frame so that only points located inside the quadrangle of the outer borders of the document remain; matching the descriptors of interest points of the current and previous frames; application of an algorithm for estimating the parameters of projective transformation between the frames; projective transformation of the quadrangle of the outer boundaries of the document from the previous frame to obtain the outer boundaries of the document in the current frame; document image normalization; calculating the color saturation and hue; updating the saturation and hue values; further considering the pixels of the normalized document image with brightness values not exceeding a preset threshold; filtration of the obtained image.

To make it possible to extract character information with a high accuracy even from a document image obtained by reading a document in which a logo mark or the like overlaps a character portion. By performing binarization processing for a document image obtained by reading a document, a binary image including first pixels representing a color darker than a reference and second pixels representing a color paler than the reference is generated. Then, by changing the pixel among the first pixels included in the generated binary image, whose corresponding pixel's color in the document image is different from a color of a character object within the document, to the second pixel, a binary image in which a background object that overlaps the character object in the document image is removed is generated.

A method of training a neural network for classifying an image into one of a plurality of classes, the method comprising: extracting, from the neural network, a plurality of subclass center vectors for each class; inputting an image into the neural network, wherein the image is associated with a predetermined class; generating, using the neural network, an embedding vector corresponding to the input image; determining a similarity score between the embedding vector and each of the plurality of subclass center vectors; updating parameters of the neural network in dependence on a plurality of the similarity scores using an objective function; extracting a plurality of updated parameters from the neural network; and updating each subclass center vector in dependence on the extracted updated parameters.

Methods and apparatus to locate barcodes in color images are disclosed. An example method includes: obtaining a color digital image using an image sensor, the image including a plurality of pixels represented by respective ones of a plurality of luminance components and respective ones of a plurality of color components; determining, based on the luminance components, a plurality of barcode indicative characteristics for respective ones of a plurality of regions of the image; determining, based on the plurality of color components, a plurality of color content amounts for respective ones of the plurality of regions; identifying one or more regions of the plurality of regions that have their respective color content amount satisfy a first criteria, and their respective barcode indicative characteristic satisfy a second criteria; and processing image data corresponding to each of the one or more regions to attempt to identify one or more barcodes in the image.

An approach is disclosed that determines an amount of time before a webpage is ready to use by a user by performing various actions. The approach captures a recording of the webpage from an invocation of the webpage for a period of time sufficient to load completely load the webpage with the capturing resulting in sequenced image frames. An AI system provides a loading point in the sequenced image frames based on an analysis of the frames input to the trained AI system. Image diversity and saturation measurements are calculated on consecutive image frames from the sequenced image frames resulting in an image change analysis. Native webpage events and times are detected from webpage characteristics gathered from the captured digital recording. The amount of time is then calculated based on the loading point from the AI system, the image change analysis; and the webpage events and their corresponding times.

A license plate number recognition method includes: extracting license plate number features of an image to be recognized including a license plate number, through a pre-trained convolutional neural network; extracting an intermediate convolution result during extracting the license plate number features, and extracting a first verification feature and/or a second verification feature according to the intermediate convolution result; verifying whether the license plate number features are correct according to the first and/or second verification features; if correct, outputting a predicted license plate number result according to the license plate number features. During the feature extraction process of the license plate number features, an intermediate feature is extracted as a verification feature to verify whether the extracted license plate number features are correct, and only when the verification is passed, outputting the license plate number result, which reduces the output error rate of the license plate number recognition result.

An image processing method, applied to an electronic apparatus, includes performing feature extraction processing on a source image to obtain a feature data set corresponding to a source file, determining an image processing parameter corresponding to the feature data set according to the feature data set, and processing the source image according to the image processing parameter.

In some forms containing keywords and content, there may be nested levels of keywords, also referred to as a hierarchy. Content in the forms may be associated with one or more keywords in one or more of the nested levels, or in the hierarchy. Identifying keywords in adjacent cells in a table (with a nested keyword being either to the right of or below another keyword) enables distinguishing between keywords and content in filled forms, and enables correct association of content with respective keywords.

Methods, systems, and computer-program products for image enhancement include receiving an image and optionally a user request, classify the image, crop image components of the image, restore cropped image components of the image, colorized restored image components, and reconstruct the image from the colorized, restored image components and other components. The other components may include text components that are restored in a separate treatment pipeline.

A license plate number recognition method includes: extracting license plate number features of an image to be recognized including a license plate number, through a pre-trained convolutional neural network; extracting an intermediate convolution result during extracting the license plate number features, and extracting a first verification feature and/or a second verification feature according to the intermediate convolution result; verifying whether the license plate number features are correct according to the first and/or second verification features; if correct, outputting a predicted license plate number result according to the license plate number features. During the feature extraction process of the license plate number features, an intermediate feature is extracted as a verification feature to verify whether the extracted license plate number features are correct, and only when the verification is passed, outputting the license plate number result, which reduces the output error rate of the license plate number recognition result.