Abstract: With this device, an edge pair likelihood map generation unit (120) computes edge pair likelihood which denotes the plausibility that a pixel of a pair is an edge of an estimation subject site, and generates an edge pair likelihood map which denotes the edge pair likelihood for each pixel. A continuity likelihood map generation unit (130) evaluates, upon the edge pair likelihood map, the continuity for the edge pair likelihood of a pixel which is included in a region wherein the estimate subject site is presumed, and generates a continuity likelihood map which denotes an edge pair likelihood which has continuity as a candidate region of the estimate subject site. An integrated likelihood map generation unit (140) generates an integrated likelihood map which denotes the candidate region which the continuity likelihood map denotes, refined based on a predetermined condition.

Abstract: Methods and apparatus to play and control playing of media content in a web page are disclosed. In one example, a method of monitoring media content in a web page loads a web page containing a media player and media content, processes an image of the web page to determine a control associated with the media player, and monitors the media content based on the control.

Abstract: Embodiments of the present disclosure disclose methods and systems for halftone frequency detection in a halftone image. The method includes receiving a first window, a second window, and the halftone image having a predetermined resolution; assigning an active value to a max variable and a min variable associated with each pixel within a first pixel neighborhood defined by the first window based on a pixel intensity value being compared with a three-way threshold set determined based on values a plurality of predetermined statistical parameters; estimating a first frequency estimate and the second frequency estimate based on a number of pixels having active values in the max variable and the min variable respectively within a second pixel neighborhood defined by the second window; and determining the halftone frequency based on the first frequency estimate value, the second frequency estimate value, and the predetermined resolution.

Abstract: Implementations generally relate to generating compositional media content. In some implementations, a method includes receiving a plurality of photos from a user, and determining one or more composition types from the photos. The method also includes generating compositions from the selected photos based on the one or more determined composition types. The method also includes providing the one or more generated compositions to the user.

Abstract: Provided are a method and an apparatus for image processing. A scene includes a low resolution depth image and a high resolution color image. A depth image up-scaler up-scales the depth image using an interpolation so as to improve the resolution of the depth image. A color-depth similarity based filter removes noise from the up-scaled depth image using the color image and the up-scaled depth image. A boundary sharpening filter converts a boundary between objects in the up-scaled depth image into a clearer boundary. The resolution of the depth image may increase to the level of the resolution of the color image by means of the depth image up-scaler, the color-depth similarity based filter and the boundary sharpening filter, and the depth of each pixel in the depth image may be adjusted.

Abstract: Systems and methods are provided for constructing a multi-resolution image from a set of compressive measurements representing a compressed version of a single native resolution image. In one aspect, compressive sense measurements are retrieved that were generated using a compressive sensing matrix and represent a compressed version of a single native resolution image of a scene. Dimensions of a desired two-dimensional, multi-resolution image are determined, and a plurality of regions is allocated, each allocated region having a respective resolution. An expansion matrix is defined to map the resolution of each allocated region to the native resolution, and a multi-resolution image of the scene is constructed using the compressive sense measurements, the compressive sensing matrix, and the defined expansion matrices. In various aspects, a full resolution or other multi-resolution images may be constructed from the same compressive measurements without regenerating new compressive measurements.

Abstract: A system and method for active contour segmentation. An image of a first and at least a second structure, a first position for the first structure, and at least a second position for the at least second structure are received, the first and the at least second positions are set as a first and an at least second initial contour, and the first and the at least second initial contours are concurrently and iteratively deformed to respectively expand into a first and at least a second expanded contour matching a shape of the first and the at least second structure by applying constraints to each point of the first and the at least second initial contour, comprising a constraint for preventing the first and at least one of the at least second initial contour from intersecting one another upon being deformed, and the constraints are updated after each iteration.

Abstract: Embodiments of the invention automatically assign colors to topics in a topic database based on a respective topic's unique identifier within the topic database. One or more functions are applied to transform each topic's unique identifier into a respective color, for example by hashing the unique identifier string, and using the hash to index into an array of colors. Thus, content items pertaining to the same topic can be consistently branded with the same color in visual presentations to the user.

Abstract: An image processing apparatus has an image acquisition unit that acquires an image that is to be subjected to processing, a learning sample extraction unit that extracts data of a plurality of learning samples from the image, a classifier learning unit that performs learning of a plurality of classifiers using the plurality of learning samples, a strong classifier generation unit that generates a strong classifier by combining the plurality of learned classifiers, and a saliency map generation unit that generates a saliency map of the image using the strong classifier.

Abstract: A method for controlling a computer system includes acquiring video of a subject, and obtaining from the video a time-resolved sequence of depth maps. An area targeting motion is selected from each depth map in the sequence. Then, a section of the depth map bounded by the area and lying in front of a plane is selected. This section of the depth map is used for fitting a geometric model of the subject.

Abstract: A document authentication method determines the authenticity of a target hardcopy document, which purports to be a true copy of an original hardcopy document. The method compares a binarized image of the target document with a binarized image of the original document which has been stored in a storage device. The image of the original document is generated by binarizing a scanned grayscale image of the original document. Halftone and non-halftone text areas in the grayscale image area separated, and the two types of text are separately binarized. The non-halftone text areas are then down-sampled. During authenticating, a scanned grayscale image of the target document is binarized by separating halftone and non-halftone text areas and binarizing them separately, and then down-sampling the non-halftone text areas. The binarized images of the target document and the original document are compared to determine the authenticity of the target document.

Abstract: Systems and methods are disclosed for controlling image annotation. One method includes acquiring a digital representation of image data and generating a set of image annotations for the digital representation of the image data. The method also may include determining an association between members of the set of image annotations and generating one or more groups of members based on the association. A representative annotation from the one or more groups may also be determined, presented for selection, and the selection may be recorded in memory.

Abstract: A method is disclosed to automatically segment 3D and higher-dimensional images into two subsets without user intervention, with no topological restriction on the solution, and in such a way that the solution is an optimal in a precisely defined optimization criterion, including an exactly defined degree of smoothness. A minimum-cut algorithm is used on a graph devised so that the optimization criterion translates into the minimization of the graph cut. The minimum cut thus found is interpreted as the segmentation with desired property.

Abstract: The present disclosure relates to an apparatus and a method for recognizing media and a financial device. The apparatus for recognizing media according to an exemplary embodiment of the present disclosure comprises: an image sensor; a skew detector configured to detect skew of the media from media image acquired by the image sensor; a noise detector configured to detect a linear component in a region of interest which include character of the media image and a noise included in the media image based on direction information of the detected linear component and the skew of the media; and a controller configured to remove the detected noise from the region of interest and recognize the character of the region of interest from which the noise is removed.

Abstract: An image processing apparatus is provided. The image processing apparatus includes an image quality processor and a controller. The image quality processor performs a detail enhancement function on an image frame. The controller calculates blur information and/or gray information of the image frame and controls the image quality processor to perform the detail enhancement function on the image frame based on the calculated information.

Abstract: According to an embodiment, a medical image processing apparatus includes an extraction unit, a calculation unit, and a selection unit. The extraction unit extracts an image region having an image element value larger than a predetermined value from a first image of at least one time phase and second images of a plurality of time phases. The calculation unit calculates a feature quantity that fluctuates in accordance with motion of the image region for the first image of at least the one time phase and the second images of the time phases. The selection unit selects the first image and the second image having similar image features of the image region based on the feature quantity from among the first image of at least the one time phase and the second images of the time phases.

Abstract: An image processing system and method of operation includes: an image processing device; an image capture module for capturing a source image on the image processing device; an assessment module for detecting an edge of the source image, and measuring an edge width measure of the edge; and a retrieving module for retrieving a depth of field from the edge width measure.

Abstract: Roughly described, pictures are divided into multiple spatial sections to be encoded by multiple section encoders. To address discontinuities of compression decisions across section boundaries, the multiple section encoders encode overlapping regions in a picture.

Abstract: A processor-implemented method for sound characterization is described. In one implementation, time-frequency transform of each of a plurality of sound signals from one or more sources, the sound signals being detected by a plurality of sensing devices, is derived. One or more single-source constant-time analysis zones based at least on correlation between the time-frequency transform signals from a pair of sensing devices are detected. At least one direction of arrival for each source in the detected single source analysis zones are detected. A histogram of the estimated directions of arrival is created and an estimate of a number of the sound sources and corresponding directions of arrival are generated based at least on the histogram.

Abstract: A method and apparatus that allows a user to specify whether his web page has primarily row or column layout. Web page editor software designed in accordance with the present invention generates the internal representation of a page (such as HTML) differently depending on how the user specifies the web page. A preferred embodiment of the present invention defaults to a primarily row layout, but allows the user to specify a primarily columnar layout by checking a box displayed by the editor. The editor generates an HTML table representing the page (or each frame of the page). Within the main HTML table, the editor generates HTML subtables, depending on whether the page is laid out using primarily row or column layout. The subtables are defined in such a way that the various elements of the page lined up in either a row or column direction, thus minimizing the problems.

Abstract: An electronic document display apparatus includes a display unit and an analysis-and-setting unit including a determination unit and a first correction unit. The display unit displays an electronic document in display layout set by the analysis-and-setting unit. If the determination unit determines that an image related to a display-range text of a display range to be displayed by the display unit is not included in the display range, the first correction unit adds an image display region where the image is displayed to the display range so as to correct the display layout in such a way that the image and at least a portion of an image text related to the image are included in the display range. The analysis-and-setting unit changes a display state of the image displayed in the image display region on the basis of a predetermined condition.

Abstract: An image summarization device includes a first image summarization section that performs a first image summarization process based on a similarity between a plurality of images to acquire a first summary image sequence, a second image summarization section that performs a second image summarization process based on a target object/scene recognition process on each image among the plurality of images to acquire a second summary image sequence, and an integration processing section that performs an integration process on the first summary image sequence and the second summary image sequence, or performs an integration process on the first image summarization process and the second image summarization process to acquire an output summary image sequence.

Abstract: Deep convolutional neural networks receive local and global representations of images as inputs and learn the best representation for a particular feature through multiple convolutional and fully connected layers. A double-column neural network structure receives each of the local and global representations as two heterogeneous parallel inputs to the two columns. After some layers of transformations, the two columns are merged to form the final classifier. Additionally, features may be learned in one of the fully connected layers. The features of the images may be leveraged to boost classification accuracy of other features by learning a regularized double-column neural network.

Abstract: An example for processing patient image data workflows can temporarily serially arrange a set of semantic labeling modules in a patient image data workflow pipeline responsive to receiving an event trigger. The example can also remove the set of modules from the patient image data workflow pipeline responsive to receiving an event completion trigger.

Abstract: A candidate-region for an object-region is set in an image. In a graph including point-S corresponding the object-region, point-T corresponding to a background-region, a point corresponding to each pixel in the image, S-link connecting each pixel and point-S, T-link connecting each pixel and point-T, and N-link connecting each pair of adjacent pixels, a cost is set for each link, graph-cut is performed. Whether a pixel connected to point-S by a link is present in the graph is judged. If no pixel is present, graph-cut is performed at each stage while costs set for all the S-links connecting each pixel in the candidate-region and point-S are increased stepwise in an increment of a predetermined threshold or less until a pixel connected to point-S by a link appears. Whether each pixel in the image belongs to the object-region or the background-region is determined based on a result of graph-cut performed last.

Abstract: A digital image processing technique, such as an image warping operation, is stored in a pre-computed lookup table (LUT) prior to image processing. The LUT represents a pixel-to-pixel mapping of pixel coordinates in a source image to pixel coordinates in a destination image. For vectors containing only inlier pixels, a fast remap table is generated based on the original LUT. Each SIMD vector listed in the fast remap is indexed to the coordinates of one of the source pixels that maps to one of the destination pixels in the vector. Other SIMD vectors that contain at least one outlier pixel are listed in an outlier index. For each vector indexed in the fast remap, linear vector I/O operations are used to load the corresponding source pixels instead of using scatter/gather vector I/O load operations via the LUT. The remaining outlier pixels are processed using scatter/gather I/O operations via the LUT.

Abstract: An optical flow tracking method includes generating a random structure according to an object to be tracked in an image, where the random structure includes at least two structure blocks with fixed relative positions; tracking all structure blocks in the random structure using an optical flow tracking algorithm to obtain a position offset of the random structure; and estimating a target position of the object according to the position offset of the random structure. According to the optical flow tracking method and apparatus provided in embodiments of the present invention, characteristic points of an object in an image are obtained, and a random structure is generated according to the characteristic points, and each random structure is tracked to obtain a position offset of the random structure, so as to estimate a position of the object, thereby reducing tracking errors and effectively improving tracking precision.

Abstract: An optical sensor includes a light source to illuminate a linear polarization in a first direction, a first optical detector disposed on a path of a light illuminated from the light source and specularly reflected by an object, a first optical element to separate the light reflected by the object into a linear polarization in the first direction and a linear polarization in a second direction orthogonal to the first direction, a second optical detector to receive the linear polarization in the second direction separated by the first optical element, and a processor to obtain an amount of the light specularly reflected by the object on the basis of an output signal of the first optical detector and an output signal of the second optical detector.

Abstract: There is described herein an image segmentation technique using an iterative process. A contour, which begins with a single point that expands into a hollow shape, is iteratively deformed into a defined structure. As the contour is deformed, various constraints are applied to points along the contour to dictate its rate of change and direction of change are modified dynamically. The constraints may be modified after one or more iterations, at each point along the contour, in accordance with newly measured or determined data.

Abstract: An image summarization device includes a first image summarization section that performs a first image summarization process based on a similarity between a plurality of images to acquire a first summary image sequence, a second image summarization section that performs a second image summarization process based on a target object/scene recognition process on each image among the plurality of images to acquire a second summary image sequence, and an integration processing section that performs an integration process on the first summary image sequence and the second summary image sequence, or performs an integration process on the first image summarization process and the second image summarization process to acquire an output summary image sequence.

Abstract: Encoding and decoding a video image having a plurality of frames using a two-step quantization and coding process are disclosed. A block of a frame are encoded by identifying pixels having certain spatial characteristics, forming a second block from the block while replacing the identified pixels with a single pixel value, such as an average of the remaining original pixels. The second block is encoded, such as by transformation and quantization, and placed into a bitstream. The second block is decoded and subtracted from the original block to generate a difference block. The difference block is encoded, such as by quantization, and is placed in the bitstream. At a decoder, both blocks are decoded and combined to reconstruct the original block.

Abstract: The present invention discloses a method for detecting a salient region of a stereoscopic image, comprising: step 1) calculating flow information of each pixel separately with respect to a left-eye view and a right-eye view of the stereoscopic image; step 2) matching the flow information, to obtain a parallax map; step 3) selecting one of the left-eye view and the right-eye view, dividing it into T non-overlapping square image blocks; step 4) calculating a parallax effect value for each of the image blocks of the parallax map; step 5) for each of the image blocks of the selected one of the left-eye view and the right-eye view, calculating a central bias feature value and a spatial dissimilarity value, and multiplying the three values, to obtain a saliency value of the image block; and step 6) obtaining a saliency gray scale map of the stereoscopic image from saliency values of the image blocks.

Abstract: A method for registering an image using a similarity criterion based on mutual information. The image to be registered is compared with a plurality of reference representations of an object, each reference representation being of a plurality of homogeneous zones. The mutual information between the image to be registered and each reference representation is calculated, on the set of homogeneous zones. The registration is given by the reference representation corresponding to the highest mutual information. The method can be advantageously applied to an aircraft aided navigation by registering images obtained by a synthetic aperture radar.

Abstract: There is provided methods, apparatuses and computer program products for image segmentation in which pixel data of an image and information of a first set of pixels of the image indicative of pixels belonging to a foreground of the image are obtained. A color model of the image is estimated. To obtain a geodesic distance map at least one geodesic distance from a pixel of the first set of pixel to another pixel of the image which is not in the first set of pixels is determined. An energy function is optimized on the basis of the geodesic distance map to obtain a segmentation of the image.

Abstract: There is provided an image processing device including capturing portions that respectively capture a first image and a second image that form an image for a right eye and an image for a left eye which can be stereoscopically viewed in three dimensions, a comparison portion that compares the first image and the second image captured by the capturing portions, a determination portion that determines, based on a comparison result of the comparison portion, which of the first image and the second image is the image for the right eye and which is the image for the left eye, and an output portion that outputs each of the first image and the second image, as the image for the right eye and the image for the left eye, based on a determination result of the determination portion.

Abstract: A system and method for active contour segmentation where an image for a structure and an initial position on the image are received, a multi-scale image representation comprising successive image levels each having associated therewith a representation of the image is computed, a representation of the image at a given level having a different image resolution than that of a representation of the image at a subsequent level, a given one of the levels at which noise in the image is removed is identified, the initial position is set as a current contour and the given level as a current level, the current contour is deformed at the current level to expand into an expanded contour matching a shape of the structure, the expanded contour is set as the current contour and the subsequent level as the current level, and the steps are repeated until the last level is reached.

Abstract: Provided are an apparatus and method for tracking a camera that reconstructs a real environment in three dimensions by using reconstruction segments and a volumetric surface. The camera tracking apparatus using reconstruction segments and a volumetric surface includes a reconstruction segment division unit configured to divide three-dimensional space reconstruction segments extracted from an image acquired by a camera, a transformation matrix generation unit configured to generate a transformation matrix for at least one reconstruction segment among the reconstruction segments obtained by the reconstruction segment division unit, and a reconstruction segment connection unit configured to rotate or move the at least one reconstruction segment according to the transformation matrix generated by the reconstruction segment division unit and connect the rotated and moved reconstruction segment with another reconstruction segment.

Abstract: Embodiments of the invention relate to detecting the number, position, and orientation of objects when a plurality of recognition target objects are present in a search target image. Dictionary image data is provided, including a recognition target pattern, a plurality of feature points of the recognition target pattern, and an offset (Ox, Oy) from the coordinates at the center of the image to the coordinates of the feature point. The sizes (Rt) and directions (?t) of feature vectors for the coordinates (Tx, Ty) of a plurality of feature points in the target image are also provided. The coordinates (Fx, Fy) of a virtual center point in the target image is derived. Additional virtual center points within a radius of the coordinates (Fx, Fy) is counted. Presence of a recognition target object is recognized near the virtual center point coordinates of the search target image.

Abstract: The present invention relates to an image background modeling and foreground extraction method based on a depth image, characterized by comprising: step 1: acquiring a depth image representing a distance from objects to a camera; step 2: initiating a real-time depth background model; step 3: updating the real-time depth background model; step 4: acquiring a current depth image representing the distance from the objects to the camera; step 5: extracting a foreground image of the current depth image based on the real-time depth background model; step 6: outputting the foreground image and generating a real-time target masking image; and step 7: updating the real-time depth background model, where code block information of each pixel point in the real-time depth background model is updated according to the real-time target masking image.

Abstract: A method and a system implementing the method for detecting presence of burnt-in text in a video sequence are described. Initially, possible text candidates are separated from the background information. Individual candidates are then validated against the character features like similar font width, resolution, width height ratio, etc. These characters are further validated for the formation of words/sentences. Finally, the temporal features are used for validation for determining valid burnt-in text candidates.

Abstract: An image segmentation method includes displaying, through a display component, an image; receiving, through a user input component, a user command relative to an original designation region of the image; shrinking the original designation region according to first depth information corresponding to the image to generate a shrunk designation region of the image; generating a plurality of seeds according to the shrunk designation region and the user command; and segmenting the image according to the seeds to obtain at least one segmented region.

Abstract: Techniques for performing foreground analysis are provided. The techniques include identifying a region of interest in a video scene; detecting a static foreground object in the region of interest; and performing a foreground analysis based on tracking information to determine whether the static foreground object is abandoned or removed.

Abstract: A method for static and moving object detection employing a motion computation method based on spatio-temporal tensor formulation, a foreground and background modeling method, and a multi-cue appearance comparison method. The invention operates in the presence of shadows, illumination changes, dynamic background, and both stopped and removed objects.

Abstract: To compress a document, a number of edges present in a selected portion of the document are counted to determine whether the number of edges exceeds a threshold. When the number of edges exceeds the threshold, a pixel is selected from the portion and a set of neighboring pixels is identified for the pixel. For each neighboring pixel in a subset of the neighboring pixels, a corresponding label of the neighboring pixel is identified. A mask layer contains labels of pixels in the portion where a label of the selected pixel is biased using labels of neighboring pixels in the subset of the neighboring pixels. The selected pixel is designated to a foreground or a background layer of the document according to the label of the selected pixel. A compressed document is constructed corresponding to the document using the mask layer, the foreground layer, and the background layer.

Abstract: A new family of morphological features, referred to herein as threshold compactness features, is provided, useful for automated classification of objects, such as cells, in images. In one embodiment, one or more thresholds and/or binary masks are applied to an image, and one or more provisional objects within a cell in the image are automatically identified. The threshold compactness of the cell is computed as a function of area S of the one or more provisional objects and border length P of the one or more provisional objects. Computation of threshold compactness allows cells in an image to be distinguished and characterized. Compared to previous techniques, the methods and apparatus described herein are more robust and computationally efficient.

Abstract: A binarizing section generates binary image data representing a binary image from raster image data representing a raster image. A white-pixel ratio determining section determines, based on the binary image data, a white pixel ratio for each line of the binary image and a position of each of lines having a white pixel ratio equal to or greater than a predetermined first threshold among the lines of the binary image, and also determines a white pixel ratio for an entirety of the binary image. An image-type determining section determines that the raster image is a photographic image when the white pixel ratio of the entirety of the binary image is equal to or less than a predetermined second threshold, and that the raster image includes a text image when the lines having a white pixel ratio equal to or greater than the first threshold appear cyclically in the binary image.

Abstract: Interactive drawing recognition is described. In one example, a command is received indicating a type of drawing and a user drawing is observed. A library of drawing templates associated with the drawing type is accessed based on the command. The observed drawing is compared to the drawing templates to identify the observed drawing, and attributes are assigned to the identified drawing.

Abstract: A method and system for real time processing of a sequence of video frames. A current frame in the sequence and at least one frame in the sequence occurring prior to the current frame is analyzed. Each frame includes a two-dimensional array of pixels. The sequence of video frames is received in synchronization with a recording of the video frames in real time. The analyzing includes performing a background subtraction on the at least one frame, which determines a background image and a static region mask associated with a static region consisting of a contiguous distribution of pixels in the current frame. The static region mask identifies each pixel in the static region upon the static region mask being superimposed on the current frame. A determination is made that a persistence requirement, both a non-persistence duration requirement and a persistence duration requirement, or a combination thereof have been satisfied.

Abstract: An image processing apparatus performs sharpening processing on each of a plurality of pieces of left-eye image data, and synthesizes them in a unit of predetermined group to generate left-eye intermediate image data (1). The image processing apparatus then performs sharpening processing on each of a plurality of pieces of left-eye intermediate image data (1) generated by the synthesizing processing and synthesizes them in a unit of predetermined group to generate left-eye intermediate image data (2).

Abstract: Provided are a shoe image retrieval apparatus and method using a matching pair, which can accurately retrieve image information corresponding to an inputted image from the database and provide the retrieved image information, by normalizing a correspondence relation in consideration of geometric image transformation about the matching pair and allowing a similar image to be retrieved by applying the normalized correspondence relation. It is possible to detect optimum geometric image transformation from a matching pair between the inputted shoe image and the image stored in the database and simultaneously retrieve a plurality of objects in the inputted shoe image based on the detected geometric image transformation, thereby providing an efficient shoe retrieval service.