Abstract: A method, system, and computer program product for modifying an appearance of an anatomical structure in a medical image, e.g., rib suppression in a chest radiograph. The method includes: acquiring, using a first imaging modality, a first medical image that includes the anatomical structure; applying the first medical image to a trained image processing device to obtain a second medical image, corresponding to the first medical image, in which the appearance of the anatomical structure is modified; and outputting the second medical image. Further, the image processing device is trained using plural teacher images obtained from a second imaging modality that is different from the first imaging modality.

Abstract: A method and system for generating a classifier to classify sub-objects of an object based on a relationship between sub-objects is provided. The classification system provides training sub-objects along with the actual classification of each training sub-object. The classification system may iteratively train sub-classifiers based on feature vectors representing the features of each sub-object, the actual classification of the sub-object, and a weight associated with the sub-object. After a sub-classifier is trained, the classification system classifies the training sub-objects using the trained sub-classifier. The classification system then adjusts the classifications based on relationships between training sub-objects. The classification system assigns a weight for the sub-classifier and weight for each sub-object based on the accuracy of the adjusted classifications.

Abstract: An image signal generated by a CCD image sensor is processed by the block-generating section 28 provided in an image-signal processing section 25. A class tap and a prediction tap are thereby extracted. The class tap is output to an ADRC process section 29, and the prediction tap is output to an adaptation process section 31. The ADRC process section 29 performs an ADRC process on the input image signal, generating characteristic data. A classification process section 30 generates a class code corresponding to the characteristic data thus generated and supplies the same to an adaptation process section 31. The adaptation process section 31 reads, from a coefficient memory 32, the set of prediction coefficients which corresponds to the class code. The set of prediction coefficients and the prediction tap are applied, thereby generating all color signals, i.e., R, G and B signals, at the positions of the pixels which are to be processed.

Abstract: A vehicle restraint system has a vision-based occupant classification system for control of airbag deployment during a crash scenario. The classification system utilizes two imaging sensors which together create a stream of paired images received and stored by an occupant classification controller. A computer program product of the controller utilizes the paired images to extract disparity/range features and stereo-vision differential edge density features. Moreover, the controller extracts wavelet features from one of the two paired images. All three features or maps are classified amongst preferably seven classifications by algorithms of the computer program product producing class confidence data fed to a sensor fusion engine of the controller for processing and output of an airbag control signal input into a restraint controller of the vehicle restraint system.

Abstract: With the present invention, data continuity is used at the time of converting an input image into high-quality image data with higher quality than the input image data, to obtain processing results which are more accurate and have higher precision. A class tap extracting unit (902) and prediction tap extracting unit (906) extract a class tap and prediction tap based on data continuity of multiple perimeter pixels corresponding to a pixel of interest in the input image, detected by a data continuity detecting unit (901).

Abstract: The present invention relates to a data conversion device and a data conversion method, in which HD image data having an image quality higher than that of the original SD image data can be obtained, to a learning device therefor, to a learning method therefor, to a program therefor, and to a recording medium therefor. A data generation section 3 generates HD image data on the basis of the class of a class tap extracted from the SD image data. On the other hand, a correction level calculation section 4 detects features of the class tap based on the class tap and the class thereof, and determines the correction level for correcting the HD image data on the basis of the features. Then, a data correction section 5 corrects the HD image data output by the data generation section 3 on the basis of the correction level. The present invention can be applied to, for example, a television receiver for receiving television broadcasting signals and displaying images.

Abstract: There is provided a method of producing compressed data including the steps of receiving input data to be compressed (step 10), transforming said input data into a sum of basis functions multiplied by a corresponding set of coefficients, which may be a transformation to the frequency domain (step 12), forming a function or row of data from said set of coefficients (step 17), estimating the function using a learning process (step 21) and recording the resulting estimate (step 28) as the compressed data. Also provided is a method of estimating a function which may be used in the method of producing compressed data including inverting the function about a predetermined value (step 18) prior to using a learning process to estimate the function. Apparatus (1) for implementing the method and a software product to cause a processing means to implement the method are also claimed.

Abstract: Each video segment in a plurality of video segments is annotated with an indicator of the likelihood that the respective video segment shows a particular feature. The plurality of video segments forms an episode of interest from a given video domain. Initial feature probabilities are calculated for respective ones of the plurality of video segments using a machine learning algorithm. Each initial feature probability indicates the likelihood that its respective video segment shows the particular feature. Refined feature probabilities are determined for respective ones of the plurality of video segments by finding the most probable state sequence in a finite state machine. This is accomplished at least in part using the determined initial feature probabilities. Finally, each of the video segments in the plurality of vides segments is annotated with its respective refined feature probability.

Abstract: An approach to clustering a set of images based on similarity measures employs a fuzzy clustering paradigm in which each image is represented by a node in a graph. The graph is ultimately partitioned into subgraphs, each of which represent true clusters among which the various images are distributed. The partitioning is performed in a series of stages by identifying one true cluster at each stage, and removing the nodes belonging to each identified true cluster from further consideration so that the remaining, unclustered nodes may then be grouped. At the beginning of each such stage, the nodes that remain to be clustered are treated as all belonging to a single candidate cluster. Nodes are removed from this single candidate cluster in accordance with similarity and connectivity criteria, to arrive at a true cluster. The member nodes of this true cluster are then removed from further consideration, prior to the next stage in the process.

Abstract: An apparatus for generating a three-dimensional representation from a two-dimensional image has a memory device for storing information for processing a two-dimensional image and for generating a three-dimensional image from the two-dimensional image, a processing device for processing a digital representation of an image by generating a two-dimensional image from the digital representation and by generating a three-dimensional image corresponding to the two-dimensional image, and an output device for outputting a three-dimensional image and a digital signal representation of the three-dimensional image. An associated method is also disclosed.

Abstract: A Support Vector Machine (110) with a Q-Metric kernel function computer (112) is provided. The Support Vector Machine (110) exhibits improved performance for classification and regression. Pattern recognition systems (100,900) that use the Support Vector Machine (110) are also provided. A Differential Evolution method of training a Support Vector Machine is also provided.

Abstract: A feature selection technique for support vector machine (SVM) classification makes use of fast Newton method that suppresses input space features for a linear programming formulation of a linear SVM classifier, or suppresses kernel functions for a linear programming formulation of a nonlinear SVM classifier. The techniques may be implemented with a linear equation solver, without the need for specialized linear programming packages. The feature selection technique may be applicable to linear or nonlinear SVM classifiers. The technique may involve defining a linear programming formulation of a SVM classifier, solving an exterior penalty function of a dual of the linear programming formulation to produce a solution to the SVM classifier using a Newton method, and selecting an input set for the SVM classifier based on the solution.

Abstract: With the present invention, data continuity is used at the time of converting an input image into high-quality image data with higher quality than the input image data, to obtain processing results which are more accurate and have higher precision. A class tap extracting unit (902) and prediction tap extracting unit (906) extract a class tap and prediction tap based on data continuity of multiple perimeter pixels corresponding to a pixel of interest in the input image, detected by a data continuity detecting unit (901).

Abstract: A method and system for creating knowledge and selecting features in a supervised classifier is disclosed. The method and system comprises changing a feature space of a plurality of defects and marking at least a portion of the samples of the defects in the feature space. The method and system includes labeling the at least a portion of the samples as training samples, determining if the training samples are of the same type and creating knowledge based upon the training samples if the samples are of the same type.

Abstract: A Lagrangian support vector machine solves problems having massive data sets (e.g.

Abstract: A processor architecture for a learning machine is presented which uses a massive array of processing elements having local, recurrent connections to form global associations between functions defined on manifolds. Associations between these functions provide the basis for learning cause-and-effect relationships involving vision, audition, tactile sensation and kinetic motion. Two arbitrary images hold each other in place in a manifold association processor and form the basis of short-term memory.

Abstract: A method and device having instructions for analyzing input data-space by learning classifiers include choosing a candidate subset from a predetermined training data-set that is used to analyze the input data-space. Candidates are temporarily added from the candidate subset to an expansion set to generate a new kernel space for the input data-space by predetermined repeated evaluations of leave-one-out errors for the candidates added to the expansion set. This is followed by removing the candidates temporarily added to the expansion set after the leave-one-out error evaluations are performed, and selecting the candidates to be permanently added to the expansion set based on the leave-one-out errors of the candidates temporarily added to the expansion set to determine the one or more classifiers.

Abstract: Judgment means for judging which of a plurality of image types, predefined by one or more items out of radiographed parts, radiography directions, and radiography methods a target image belongs to, according to characteristic quantities of the target image is generated and prepared through machine learning using sample images belonging to each of the image types, and an image type of a radiograph included in an input image is judged by applying the judgment means to the radiograph.

Abstract: There is disclosed a system and method for interpreting and describing graphic images. In an embodiment, the method of inserting a description of an image into an audio recording includes: interpreting an image and producing a word description of the image including at least one image keyword; parsing an audio recording into a plurality of audio clips, and producing a transcription of each audio clip, each audio clip transcription including at least one audio keyword; calculating a similarity distance between the at least one image keyword and the at least one audio keyword of each audio clip; and selecting the audio clip transcription having a shortest similarity distance to the at least one image keyword as a location to insert the word description of the image. The word description of the image can then be appended to the selected audio clip to produce an augmented audio recording including the interpreted word description of the image.

Abstract: Given an image of structured and/or unstructured objects we automatically partition it into semantically meaningful areas each labeled with a specific object class. We use a novel type of feature which we refer to as a shape filter. Shape filters enable us to capture some or all of shape, texture and appearance context information. A shape filter comprises one or more regions of arbitrary shape, size and position within a bounding area of an image, paired with a specified texton. A texton comprises information describing the texture of a patch of surface of an object. In a training process we select a sub-set of possible shape filters and incorporate those into a conditional random field model of object classes. That model is then used for object detection and recognition.

Abstract: A method and system for generating a detector to detect a dominant color of an image is provided. A dominant color system trains a detector to classify colors as being dominant colors of images. The dominant color system trains the detector using a collection of training images. To train the detector, the dominant color system first identifies candidate dominant colors of the training images. The dominant color system then extracts features of the candidate dominant colors. The dominant color system also inputs an indication of dominance of each of the candidate dominant colors. The dominant color system then trains a detector to detect the dominant color of images using the extracted features and indications of dominance of the candidate dominant colors as training data.

Abstract: It is to learn an object identification parameter while suppressing an influence of the background area. The object identification parameter learning system includes: a feature extracting device for obtaining a feature of an object from the image; a background specifying device for specifying a background area of the image; a background replacing device which replaces feature components corresponding to the background area of the feature with other values; and an identification parameter update device for updating the identification parameter based on the feature components replaced by the background replacing device. The identification parameter can be learnt by generating a plurality of pieces of feature data of the object with different backgrounds from a single object image through replacing the background area of the feature of the object.

Abstract: Designs for cognitive memory systems storing input data, images, or patterns, and retrieving it without knowledge of where stored when cognitive memory is prompted by query pattern that is related to sought stored pattern. Retrieval system of cognitive memory uses autoassociative neural networks and techniques for pre-processing query pattern to establish relationship between query pattern and sought stored pattern, to locate sought pattern, and to retrieve it and ancillary data.

Abstract: A technique is provided for collaboratively processing and/or analyzing a set of image data. The technique provides for the initiation of a collaborative session by an application server. One or more collaborative workstations may join the collaborative session, providing common access to an image data set and to tools for processing and/or viewing the image data set. Operators at the collaborative workstations and/or the application server may thereby simultaneously access, process, and/or analyze the image data set. Communication between the operators via the network supporting the collaborative session may also be provided.

Abstract: An architecture for robot intelligence enables a robot to learn new behaviors and create new behavior sequences autonomously and interact with a dynamically changing environment. Sensory information is mapped onto a Sensory Ego-Sphere (SES) that rapidly identifies important changes in the environment and functions much like short term memory. Behaviors are stored in a DBAM that creates an active map from the robot's current state to a goal state and functions much like long term memory. A dream state converts recent activities stored in the SES and creates or modifies behaviors in the DBAM.

Abstract: Various technologies and techniques are disclosed for improving handwriting recognition using a neural network by allowing a user to provide samples. A recognition operation is performed on the user's handwritten input, and the user is not satisfied with the recognition result. The user selects an option to train the neural network on one or more characters to improve the recognition results. The user is prompted to specify samples for the certain character, word, or phrase, and the neural network is adjusted for the certain character, word, or phrase. Handwritten input is later received from the user. A recognition operation is performed on the handwritten input using the neural network that was adjusted for the certain character or characters.

Abstract: A computerized directed feature development method receives an initial feature list, a learning image and object masks. Interactive feature enhancement is performed by human to generate feature recipe. The Interactive feature enhancement includes a visual profiling selection method and a contrast boosting method. A visual profiling selection method for computerized directed feature development receives initial feature list, initial features, learning image and object masks. Information measurement is performed to generate information scores. Ranking of the initial feature list is performed to generate a ranked feature list. Human selection is performed through a user interface to generate a profiling feature. A contrast boosting feature optimization method performs extreme example specification by human to generate updated montage. Extreme directed feature ranking is performed to generate extreme ranked features.

Abstract: This invention overcomes the disadvantages of the prior art by providing a human/machine interface (HMI) for use with machine vision systems (MVSs) that provides the machine vision system processing functionality at the sensor end of the system, and uses a communication interface to exchange control, image and analysis information with a standardized, preferably portable device that can be removed from the MVS during runtime. In an illustrative embodiment, this portable device can be a web-browser equipped computer (handheld, laptop or fixed PC) or a Personal Digital Assistant (PDA). The communication interface on the sensor-end of the system is adapted to communicate over a cable or wireless communication link (for example infrared (IR) or radio frequency (RF)), with a corresponding communication interface in the portable device.

Abstract: This invention provides an information fusion method for multiple indicators of cancers detected with multiple channels. Each channel consists of specifically tuned detectors, features, classifiers and Bayes networks. The outputs of the Bayes networks are probabilities of malignancy for the detections passing the corresponding classifier.

Abstract: JPEG encoded data are entropy-decoded to quantized DCT coefficients which are sent to a prediction tap extraction circuit (41) and to a class tap extraction circuit (42). The prediction tap extraction circuit (41) and the class tap extraction circuit (42) extract what is needed from the quantized DCT coefficients to form prediction taps and class taps. A classification circuit (43) effects classification based on the class taps. A coefficient table storage unit (44) sends tap coefficients corresponding to the classes resulting from the classification to a sum of products circuit (45), which sum of products circuit (45) then effects linear predictive calculations, using the tap coefficients and the class taps, to generate decoded picture data.

Abstract: A system and method facilitating pattern recognition is provided. The invention includes a pattern recognition system having a convolutional neural network employing feature extraction layer(s) and classifier layer(s). The feature extraction layer(s) comprises convolutional layers and the classifier layer(s) comprises fully connected layers. The pattern recognition system can be trained utilizing a calculated cross entropy error. The calculated cross entropy error is utilized to update trainable parameters of the pattern recognition system.

Abstract: The present invention provides a method and apparatus for accelerated handwritten symbol recognition in a pen based tablet computer. In one embodiment, handwritten symbols are translated into machine readable characters using special purpose hardware. In one embodiment, the special purpose hardware is a recognition processing unit (RPU) which performs feature extraction and recognition. A user inputs the handwritten symbols and software recognition engine preprocesses the input to a reduced form. The data from the preprocessor is sent to the RPU which performs feature extraction and recognition. In one embodiment, the RPU has memory and the RPU operates on data in its memory. In one embodiment, the RPU uses a hidden Markov model (HMM) as a finite state machine that assigns probabilities to a symbol state based on the preprocessed data from the handwritten symbol. In another embodiment, the RPU recognizes collections of symbols, termed “wordlets,” in addition to individual symbols.

Abstract: Embodiments of the present invention comprise methods and systems for automatically adjusting images to conform to preference data.

Abstract: The invention provides a method for identifying or quantifying characteristics of interest of unknown objects, comprising training a single neural network model with training sets of known objects having known values for the characteristics; validating the optimal neural network model; and analyzing unknown objects having unknown values of the characteristics by imaging them to obtain a digital image comprising pixels representing the unknown objects, background and any debris; processing the image to identify, separate, and retain pixels representing the unknown objects from pixels and to eliminate background and debris; analyzing the pixels representing each of the unknown objects to generate data representative of image parameters; providing the data to the flash code deployed from the candidate neural network model; analyzing the data through the flash code; and receiving output data (the unknown values of the characteristics of interest of the unknown objects) from the flash code in a predetermined forma

Abstract: A novel apparatus and method for controlling a system with multiple observable variables is disclosed. The apparatus and method disclosed use elements of the bottom-up and top-down strategies of artificial intelligence to provide a control system that is able to learn without a training set of information, and that has an learning process that can handle any amount of input data, i.e., cannot become saturated. The control system disclosed is capable of learning and controlling non linear dynamical systems. The control system is also capable of adding additional observable variables or subtracting existing observable variables to determine the state of the plant or system being controlled.

Abstract: A segmentation method receives a learning image and an objects of interest specification. A segmentation learning method creates a segmentation recipe output. It performs a segmentation application using the second image and the segmentation recipe to create a segmentation result output. The segmentation learning method includes an object region of interest segmentation learning step and an object type specific segmentation learning step. The segmentation application method includes an object region of interest segmentation step and an object type specific segmentation step. The learnable object segmentation method further comprises an online learning and a feedback learning step that allows the update of the segmentation recipe automatically or under user direction.

Abstract: A workpiece W serving as an object for detection is fixed in place. A camera 20 is mounted to the end of a robot RB. The camera is turned about an axis, which passes through the center position of the workpiece W and is perpendicular to the optical axis of the camera, to take an image of the workpiece W at a plurality of positions in different directions. A teaching model is generated on the basis of each produced image data. The relative position and posture of the workpiece to the camera 20 is also stored in association with the teaching model. Thus, it is possible to easily generate the teaching model of the workpiece regardless of three-dimensional variations in posture.

Abstract: In a method and system for retrieving a medical image from a data base, an image processing algorithm is applied to a first medical image. The result of the applied image processing algorithm to the first medical image is compared with a plurality of results being stored in a data base. Each result of the plurality of results is the result of the image-processing algorithm being applied to each medical image of a plurality of medical images which is stored in the data base. Then, a second medical image belonging to the plurality of medical images is determined and retrieved from the data base, wherein the relevant result is comparable to the relevant result related to the first medical image.

Abstract: A procedure for fast training and evaluation of support vector machines (SVMs) with linear input features of high dimensionality is presented. The linear input features are derived from raw input data by means of a set of m linear functions defined on the k-dimensional raw input data. Training uses a one-time precomputation on the linear transform matrix in order to allow training on an equivalent training set with vector size k instead of m, given a great computational benefit in case of m>>k. A similar precomputation is used during evaluation of SVMs, so that the raw input data vector can be used instead of the derived linear feature vector.

Abstract: Teaching data setting apparatus and a method for image processing are provided to enable setting optimum teaching data to achieve reduction in errors and error detection in the image processing. A subject teaching data generating section 1 uses databases of subject attributes and recognition operation conditions for a plurality of pieces of teaching data representing all possible combinations of the subject attributes and the recognition operation conditions. A teaching data candidate selecting section 2 selects, from the generated plurality of pieces of teaching data, teaching data candidates related to a subject to be mounted and a mounting machine for use. A subject image input section 3 is supplied with a simple subject image representing only the subject. An evaluation image generating section 4 generates predetermined evaluation images by using the simple subject image.

Abstract: An adaptive interface for a programmable system, for predicting a desired user function, based on user history, as well as machine internal status and context. The apparatus receives an input from the user and other data. A predicted input is presented for confirmation by the user, and the predictive mechanism is updated based on this feedback. Also provided is a pattern recognition system for a multimedia device, wherein a user input is matched to a video stream on a conceptual basis, allowing inexact programming of a multimedia device. The system analyzes a data stream for correspondence with a data pattern for processing and storage. The data stream is subjected to adaptive pattern recognition to extract features of interest to provide a highly compressed representation which may be efficiently processed to determine correspondence.

Abstract: A method and an apparatus operate like a human neural network to analyze and store input information and form patterns according to the input and stored information. The apparatus has a preprocessing unit (3), an activity computation unit (5), a mutual repression unit (6), and a composition unit (7). The apparatus receives an input pattern, calculates the similarity and activity levels of each stored pattern with respect to the input pattern, and repeats a predetermined number of times the activity calculation of each stored pattern according to the calculated activity level (A(i)), a negative repression coefficient, and the activity levels of the other stored patterns. The apparatus applies final activity levels to cell values of the stored patterns, totals the cell values through the stored patterns, and generates a new pattern according to the totaled cell values.

Abstract: A text and imagery spatial correlator automatically relates the geographic locations where events referenced in text occur to those same geographic locations appearing in imagery. In the preferred embodiment, the system deploys adaptable, context-sensitive agents for the automatic interpretation of text, and the application of those agents to classify and geolocate textual references. The results are integrated with similar spatial references from imagery in a common data model, accessible by both spatial and non-spatial data correlation. The unique combination of elements enables the system to automatically develop spatial relationships between text and imagery. Although the invention finds utility in numerous applications, the system and method are particularly useful in correlating remote sensor imagery.

Abstract: A method to arbitrate between results obtained from observationally derived data by several procedures, where the results are items characteristic of the observationally derived data is disclosed. Each procedure is given a ranking according to the confidence on the procedure. The results and characteristics derived from the results are used in a plurality of rules, where the rules are used to arbitrate between the results.

Abstract: A data processing apparatus processes input data and outputs the processed data. The data processing apparatus includes a data processing section and a real-time learning section. The data processing section processes the input data by a predetermined processing method and outputs the processed data. The real-time learning section controls such that the processing method is learned in real time and the data processing section processes the input data by the learned processing method, so that the output data is improved as time elapses.

Abstract: A method of training a user to identify an object in an image, by querying a computer system having a lexicon of photo-interpreters to formulate object extraction rules. A database to be queried by an expert photo analyst is generated and a programming language is provided to generate extraction rules against that database. Graphical query results are displayed to determine whether an object has been identified. First, a hyperspectral image cube is represented by a set of fraction planes and texture transforms. Mean spectral reading values are obtained and used to build a pseudo multivariate distribution of the values. Recognizable features are then extracted from the hyperspectral image cube. A textural transform of a hyperspectral image cube results. The hundreds of image bands reduced to a few texture transforms can be used to train users of the image recognition system.

Abstract: A system and method facilitating pattern recognition is provided. The invention includes a pattern recognition system having a convolutional neural network employing feature extraction layer(s) and classifier layer(s). The feature extraction layer(s) comprises convolutional layers and the classifier layer(s) comprises fully connected layers. The pattern recognition system can be trained utilizing a calculated cross entropy error. The calculated cross entropy error is utilized to update trainable parameters of the pattern recognition system.

Abstract: A method of effacing zipper image, applicable to scan a document by a scanner with a stagger charge-coupled device. The nth (n is a positive integer) pixel obtained by scan on a scan line at which the zipper image is produced is subtracted by the (n+1)th pixel obtained by scan on the scan line. The absolute value of the result is then compared to a critical value. If the result is smaller than the critical value, the nth pixel is the nth pixel modified as the nth pixel after process.

Abstract: A method for combining a random set of video features non-linearly to evaluate perceptual quality of video sequences includes (a) receiving a video sequence for image quality evaluation; (b) providing an objective metric image quality controller comprising a random set of metrics ranging from M1 to Mn without dependency information for each one metric; (c) applying each one metric individually to the video sequence to provide an individual objective scoring value of the video sequence ranging from x1 to xn; (d) determining a plurality of sets of weights (w1 to wn) which correlate to predetermined subjective evaluations of image quality for a predetermined plurality of video sequences (n), each one set of weights being assigned a range having an incremental value equal to the range divided by a number of combinations for each one set of weights; (e) weighting each individual objective scoring value x1 to xn provided by each one metric of the random set of metrics in step (c); (f) combining metrics of the weigh

Abstract: A system and method for selecting a best match of a received input signal from a set of candidate signals, wherein two or more of the candidate signals are uncorrelated. In a preprocessing phase a unified signal transform (UST) is determined from the candidate signals. The UST converts each candidate signal to a generalized frequency domain. The UST is applied at a generalized frequency to each candidate signal to calculate corresponding generalized frequency component values (GFCVs) for each candidate signal. At runtime, the input signal of interest is received, and the UST is applied at the generalized frequency to the input signal of interest to calculate a corresponding GFCV. The best match is determined between the GFCV of the input signal of interest and the GFCVs of each of the set of candidate signals. Finally, information indicating the best match candidate signal from the set of candidate signals is output.