Abstract: A method and system for obstacle mapping for navigation of an autonomous vehicle is disclosed. The method comprises providing an autonomous vehicle with an image capturing device, and focusing the image capturing device at a predetermined number of different specified distances to capture an image at each of the specified distances. The method further comprises identifying which regions in the captured images are in focus, and assigning a corresponding lens-focus distance to each of the regions that are in focus. A composite image is formed from the captured images, with each of the regions labeled with the corresponding lens-focus distance. A three-dimensional obstacle map is then produced from the composite image. The three-dimensional obstacle map has an x, y, z coordinate system, with x being proportional to pixel horizontal position, y being proportional to pixel vertical position, and z being the lens-focus distance.

Abstract: The present invention provides a method of detecting borders in a chest image. The method comprises calculating the gray level difference between neighboring pixels of the chest image to get a differential image; filtering the differential image to reduce noise and enhance the target; binarizing the filtered image; picking out the outside edges of the area with high gray level from the binarized image; and determining a final border from the edges according to the character of the border to be found. The present invention also provides an apparatus and a storage medium for implementing the above-mentioned method and a method and an apparatus for automatically measuring the Cardiothoracic Ratio of a chest image.

Abstract: An efficient, effective and at times superior object detection and/or recognition (ODR) function may be built from a set of Bayesian stumps. Bayesian stumps may be constructed for each feature and object class, and the ODR function may be constructed from the subset of Bayesian stumps that minimize Bayesian error for a particular object class. That is, Bayesian error may be utilized as a feature selection measure for the ODR function. Furthermore, Bayesian stumps may be efficiently implemented as lookup tables with entries corresponding to unequal intervals of feature histograms. Interval widths and entry values may be determined so as to minimize Bayesian error, yielding Bayesian stumps that are optimal in this respect.

Abstract: A method of detecting a road feature in an image signal derived from an infrared-sensitive camera. The method, in overview, comprises processing an image frame by assigning binary values to pixels in the frame in response to their representative temperature, and then to analyze spatially the binary mask to identify regions of pixels having mutually similar assigned binary values. The road feature is subsequently found from the analysis of the identified regions of mutually similar binary values and a visual indication of the road feature in relation to the image frame provided to the user.

Abstract: A calculation processing apparatus for executing network calculations defined by hierarchically connecting a plurality of logical processing nodes that apply calculation processing to input data, sequentially designates a processing node which is to execute calculation processing based on sequence information that specifies an execution order of calculations of predetermined processing units to be executed by the plurality of processing nodes, so as to implement the network calculations, and executes the calculation processing of the designated processing node in the processing unit to obtain a calculation result. The calculation apparatus allocates partial areas of a memory to the plurality of processing nodes as ring buffers, and writes the calculation result in the memory while circulating a write destination of data to have a memory area corresponding to the amount of the calculation result of the processing unit as a unit.

Abstract: A unique multi-stage classification system and method that facilitates reducing human resources or costs associated with text classification while still obtaining a desired level of accuracy is provided. The multi-stage classification system and method involve a pattern-based classifier and a machine learning classifier. The pattern-based classifier is trained on discriminative patterns as identified by humans rather than machines which allow a smaller training set to be employed. Given humans' superior abilities to reason over text, discriminative patterns can be more accurately and more readily identified by them. Unlabeled items can be initially processed by the pattern-based classifier and if no pattern match exists, then the unlabeled data can be processed by the machine learning classifier. By employing the classifiers in this manner, less human involvement is required in the classification process. Even more, classification accuracy is maintained and/or improved.

Abstract: The present invention extends to methods, systems, and computer program products for recognizing input gestures. A neural network is trained using example inputs and backpropagation to recognize specified input patterns. Input gesture data is representative of movements in contact on a multi-touch input display surface relative to one or more axes over time. Example inputs used for training the neural network to recognize a specified input pattern can be created from sampling input gesture data for example input gestures known to represent the specified input pattern. Trained neural networks can subsequently be used to recognize input gestures that are similar to known input gestures as the specified input pattern corresponding to the known input gestures.

Abstract: A crystal lookup table used to define a matching relationship between a signal position of a detected event in a PET scanner and a corresponding detector pixel location is generated using a neural network-based algorithm, and is implemented by a FPGA.

Abstract: The disclosed terrain model is a generative, probabilistic approach to modeling terrain that exploits the 3D spatial structure inherent in outdoor domains and an array of noisy but abundant sensor data to simultaneously estimate ground height, vegetation height and classify obstacles and other areas of interest, even in dense non-penetrable vegetation. Joint inference of ground height, class height and class identity over the whole model results in more accurate estimation of each quantity. Vertical spatial constraints are imposed on voxels within a column via a hidden semi-Markov model. Horizontal spatial constraints are enforced on neighboring columns of voxels via two interacting Markov random fields and a latent variable. Because of the rules governing abstracts, this abstract should not be used to construe the claims.

Abstract: Various technologies and techniques are disclosed that improve handwriting recognition operations. Handwritten input is received in training mode and run through several base recognizers to generate several alternate lists. The alternate lists are unioned together into a combined alternate list. If the correct result is in the combined list, each correct/incorrect alternate pair is used to generate training patterns. The weights associated with the alternate pairs are stored. At runtime, the combined alternate list is generated just as training time. The trained comparator-net can be used to compare any two alternates in the combined list. A template matching base recognizer is used with one or more neural network base recognizers to improve recognition operations. The system provides comparator-net and reorder-net processes trained on print and cursive data, and ones that have been trained on cursive-only data. The respective comparator-net and reorder-net processes are used accordingly.

Abstract: A method for acquiring an image on an imaging system includes accessing at least first image data from a first imaging system, processing the first image data in accordance with a CAD algorithm, acquiring at least second image data based upon results of the CAD algorithm and processing the second image data in accordance with the CAD algorithm to confirm the results of the CAD algorithm regarding the first image data.

Abstract: A convolutional neural network is implemented on a graphics processing unit. The network is then trained through a series of forward and backward passes, with convolutional kernels and bias matrices modified on each backward pass according to a gradient of an error function. The implementation takes advantage of parallel processing capabilities of pixel shader units on a GPU, and utilizes a set of start-to-finish formulas to program the computations on the pixel shaders. Input and output to the program is done through textures, and a multi-pass summation process is used when sums are needed across pixel shader unit registers.

Abstract: The normal vector information generating device generates normal vector information on the surface of an object, and includes: an image information obtaining unit obtaining information about an image of the object, the information including luminance information about luminance of light from the object and polarization information about polarization of the light; a shadow area extracting unit extracting an attached shadow area and a cast shadow area from the image based on the luminance information and the polarization information, the attached shadow area appearing on the surface of the object depending on an angle of incidence light, and the cast shadow area appearing on the surface of a material body when the light is blocked by the object; and a normal vector information generating unit generating normal vector information identifying a normal vector on the surface of the object in the attached shadow area using the polarization information.

Abstract: A probabilistic boosting tree framework for computing two-class and multi-class discriminative models is disclosed. In the learning stage, the probabilistic boosting tree (PBT) automatically constructs a tree in which each node combines a number of weak classifiers (e.g., evidence, knowledge) into a strong classifier or conditional posterior probability. The PBT approaches the target posterior distribution by data augmentation (e.g., tree expansion) through a divide-and-conquer strategy. In the testing stage, the conditional probability is computed at each tree node based on the learned classifier which guides the probability propagation in its sub-trees. The top node of the tree therefore outputs the overall posterior probability by integrating the probabilities gathered from its sub-trees. In the training stage, a tree is recursively constructed in which each tree node is a strong classifier. The input training set is divided into two new sets, left and right ones, according to the learned classifier.

Abstract: The present invention provides a method and apparatus for appropriately calculating 3D volume of cerebral hemorrhage lesion. The present invention sets a ROI for normal cerebral parenchymal region and a ROI for cerebral hemorrhage lesion respectively for a plurality of continuous slice images obtained by imaging the head having cerebral hemorrhage onset with an X-ray CT device, determines a mean value of the CT values of pixels within the ROI of normal cerebral parenchymal region for each of a plurality of images, then determines an amount of shift of mean values of images based on the mean value of one image, adjusts the CT value of pixels in the ROI of cerebral hemorrhage lesion for each image by using the amount of shift for each image, specifies the maximum value of the adjusted CT value through a plurality of images, and then calculates the 3D volume of the cerebral hemorrhage lesion.

Abstract: Techniques are described for detecting anomalous events using a long-term memory in a video analysis system. The long-term memory may be used to store and retrieve information learned while a video analysis system observes a stream of video frames depicting a given scene. Further, the long-term memory may be configured to detect the occurrence of anomalous events, relative to observations of other events that have occurred in the scene over time. A distance measure may used to determine a distance between an active percept (encoding an observed event depicted in the stream of video frames) and a retrieved percept (encoding a memory of previously observed events in the long-term memory). If the distance exceeds a specified threshold, the long-term memory may publish the occurrence of an anomalous event for review by users of the system.

Abstract: In a hierarchical neural network having a module structure, learning necessary for detection of a new feature class is executed by a processing module which has not finished learning yet and includes a plurality of neurons which should learn an unlearned feature class and have an undetermined receptor field structure by presenting a predetermined pattern to a data input layer. Thus, a feature class necessary for subject recognition can be learned automatically and efficiently.

Abstract: A method of processing information received in the form of a message from a mobile handheld device with a camera, where the message includes data representing an image of a document and an identifier of a type of said document. The method preferably provides image enhancement, information recognition and analysis selected based on the identified document type. An output of the system is preferably provided to the user in the form specified by the user, such as an e-mail.

Abstract: A method for computer-aided detection of microcalcification clusters obtains digital mammography data for a single view image and normalizes and filters the image data to reduce noise. A first mask is generated and applied to the image data for defining the breast structure, forming a first cropped image. A second mask is generated and applied to the image data for defining muscle structure, forming a second cropped image. An artifact mask corresponding to vascular calcifications and known imaging artifacts is generated and applied to the first and second cropped images, defining first and second artifact-masked cropped images. In a repeated sequence, portions of each artifact-masked cropped image are processed using an enhancement algorithm and reducing edge effects to obtain a set of microcalcification cluster candidates and suspected microcalcification clusters.

Abstract: The present invention discloses methods and systems for discovering printers and shares on a computer network. Each domain on the network is identified, and each computer in the domain is identified. In addition, each printer connected to the computer and each share on the computer is identified. Shortcuts to the identified printers and shares are created on at least one computer on the network. Moreover, drivers are preferably installed on the computer for each printer for which a shortcut was created. In the event that the total number of resources (i.e., shares and/or printers) exceeds a threshold, then the process terminates. Otherwise, the present invention continues until all printers and shares on the network are identified, and the appropriate shortcuts are created. Thus, the present invention provides methods and systems for discovering resources on a network.

Abstract: A system and method for processing machine learning techniques (such as neural networks) and other non-graphics applications using a graphics processing unit (GPU) to accelerate and optimize the processing. The system and method transfers an architecture that can be used for a wide variety of machine learning techniques from the CPU to the GPU. The transfer of processing to the GPU is accomplished using several novel techniques that overcome the limitations and work well within the framework of the GPU architecture. With these limitations overcome, machine learning techniques are particularly well suited for processing on the GPU because the GPU is typically much more powerful than the typical CPU. Moreover, similar to graphics processing, processing of machine learning techniques involves problems with solving non-trivial solutions and large amounts of data.

Abstract: A method and system for training a neural network of a visual recognition computer system, extracts at least one feature of an image or video frame with a feature extractor; approximates the at least one feature of the image or video frame with an auxiliary output provided in the neural network; and measures a feature difference between the extracted at least one feature of the image or video frame and the approximated at least one feature of the image or video frame with an auxiliary error calculator. A joint learner of the method and system adjusts at least one parameter of the neural network to minimize the measured feature difference.

Abstract: The picture archiving and communication system (PACS) substantially optimizes a post-scanning command process in accordance with a user-specified priority for further processing a selected set of imaging data that has been collected by medical imaging devices such as a CT scanner and a MRI scanner and stored in distributed storage units on the network. The user-specified priority includes the least response time, the least costs and the least network traffic. The user-specified priority is optionally fixed in some preferred embodiments.

Abstract: This specification discloses a system of using a neural network to distinguish text and pictures in an image and the method thereof. Using the knowledge of text recognition learned by the neural network in advance, images data of color brightness and gray levels in an image block are processed to generate a greatest text faith value. The system determines the text status of the image block by comparing a text threshold with the greatest text faith value. If the greatest text faith value is larger than the text threshold, then the image block is determined to contain text pixels; otherwise, the image block contains purely picture pixels. This achieves the goal of separating text and pictures in an image.

Abstract: A computer-readable medium for performing target object recognition in images and video includes instructions for receiving target image data including a target object, applying non-negative matrix factorization with enforced sparseness to the target image data to generate target extracted image feature data, training a neural network to identify the target object using the target extracted image feature data to obtain a trained neural network, receiving object image data, applying non-negative matrix factorization with enforced sparseness to the object image data to generate object extracted image feature data, analyzing the object extracted image feature data with the trained neural network to obtain a result indicating whether the presence of the target object is identified in the object image data, and storing the result of analyzing the object extracted image feature data.

Abstract: Th invention relates to a method and system for modifying a digital image (100) consisting of pixels. Said digital image is divided into areas (103). At least one area value is assigned to each area Zi (103). At least one parameter value Vpij (203) is assigned to each of said areas (103). A set of couples (Zi, Vpij) forms a parameter image (201). The inventive method consists (a) in determining the determined parameter values Vpir for each area (103), the parameter image (201) being called determined parameter image, (b) in adjusting the determined parameter image by reducing the variations thereof, (c) in modifying pixel values (102) of the determined pixel (101) of said digital image (100) according to the parameter values (203) of said adjusted parameter image, whereby the digital image is differentially modified for each of said pixels and quasi regularly for contiguous areas.

Abstract: An improved method and mechanism for recognizing chirographs (handwritten characters) input into a computer system. A primary recognizer is provided for converting chirographs to code points, and secondary recognizers such as binary CART trees are developed and trained to differentiate chirographs which produce certain code points at the primary recognizer. Each such secondary recognizer is associated with each selected code point. When a chirograph is received, the chirograph is provided to the primary recognizer whereby a code point corresponding thereto is received. If the code point corresponds to one of the secondary recognizers, the chirograph is passed to the secondary recognizer, and a code point is returned from the secondary recognizer. If not, the code point provided by the primary recognizer is returned.

Abstract: In a microarray image analysis system, when one of a plurality of statuses is set for a spot of a microarray by the user, the status of a similar spot is automatically determined. In a microarray image, the user determines a status of a spot, the pixel value matrix of an image in a spot region is learned by a neural network, a vertically and horizontally symmetrical image and an image rotated about the center of the region are formed and are learned by the neural network, and the neural network formed by repeating these steps is used for automatically recognizing the status of an undecided spot.

Abstract: A neural network is trained and used to reduce artifacts in spatial domain representations of images that were compressed by a transform method and then decompressed. For example, the neural network can be trained and used to reduce artifacts such as blocking and ringing artifacts in JPEG images.

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, system, and computer product is presented for mapping a set of patterns into an m-dimensional space so as to preserve relationships that may exist between these patterns. A subset of the input patterns is chosen and mapped into the m-dimensional space using an iterative nonlinear mapping process based on subset refinements. A set of n attributes are determined for each pattern, and one or more neural networks or other supervised machine learning techniques are then trained in accordance with the mapping produced by the iterative process. Additional input patterns not in the subset are mapped into the m-dimensional space by determining their n input attributes and using the neural networks in a feed-forward (prediction) mode.

Abstract: Systems and methods of robust perceptual color identification are disclosed. The methods include a multilevel analysis for determining the robust perceptual color of an object based on observed colors. This multilevel analysis can include a pixel level, a frame level, and/or a sequence level. The determination may make use of color drift matrices and trained functions such as statistical probability functions. The color drift tables and function training are based on training data generated by observing objects of known robust perceptual color in a variety of circumstances. Embodiments of the invention are applicable to the identification and tracking of objects, for example, in a surveillance video system.

Abstract: The present invention provides an image and video indexing scheme for content analysis. According to the invention, a database of images or videos is compressed. By examining patterns in the compression scheme of each image or video, the present invention identifies the content of the data. In one embodiment, an unsupervised learning method is employed where each image or video is sub-divided into smaller blocks (8 pixels×8 pixels, for instance) and each of the smaller blocks is examined for its compression pattern. Then, the patterns associated with each of the smaller blocks is recorded for each of the images in the database and content is retrieved from the database by associating certain patterns or groups of patterns with certain content.

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 an image processing apparatus that efficiently identifies input characters and the like by an intensity image analysis based on range information. The image processing apparatus performs distance measurement by performing three-dimensional measurement by pattern projection to characters written to a manuscript, whiteboard, and the like, gets an intensity image used as a so-called actual image and an image for distance measurement in parallel, and identifies input characters by an intensity image analysis based on range information. The image processing apparatus picks up characters, patterns, and the like written onto paper or the like with a pen, performs geometric transformation for picked-up images, and performs natural input character analysis and reading processing. Furthermore, by comparing images picked up in a time-series, noise elimination and manuscript position modifications become possible.

Abstract: The present invention is directed to a system, method, and apparatus for orienting images. A neural net is trained with images of known orientation and an indicator indicating such known orientation. Images of unknown orientation are then input to the neural net and the orientation is determined based on the output of the neural net.

Abstract: A method and an apparatus of designing a set of wavelet basis trained to fit a particular problem. The method and apparatus include constructing a neural network of arbitrary complexity using a discrete and finite Radon transform, feeding an input wavelet prototype through the neural network and its backpropagation to produce an output, and modifying the input wavelet prototype using the output.

Abstract: A method and apparatus of reconstructing data from higher moment data. The method and apparatus include performing a finite Radon transform, generating an average function to allow inversion of transform in one step and correlating the transform output at each point. A resultant set of duplications is calculated using the correlation process to generate a new average function. Partial backprojections of the Radon transform are summed at each point, and the new average function for each point is subtracted from the sum of the partial backprojections at that point.

Abstract: A real-time screen recording system is disclosed. Under the design of the real-time screen recording system of this invention, the Event Inspector records updated regions of the screen when a windows update event is generated, the Periodical Extractor records the screen updates periodically. This mechanism ensures that all updates will be recorded. The Event Filter inspects the update events to identify events relating to updates of screen to ensure correct recording of the updated regions. The Sporadic Controller adjusts the working frequency of the Periodical Extractor, so that the Periodical Extractor operates in an efficient manner. The Update Region Filter filters out unnecessary updated regions to avoid heavy workload of the computer system. As a result, an efficient and correct screen recording system may be achieved.

Abstract: In a microarray image analysis system, when one of a plurality of statuses is set for a spot of a microarray by the user, the status of a similar spot is automatically determined.

Abstract: A method of training an artificial neural network (ANN) involves receiving a likelihood distribution map as a teacher image, receiving a training image, moving a local window across sub-regions of the training image to obtain respective sub-region pixel sets, inputting the sub-region pixel sets to the ANN so that it provides output pixel values that are compared to output pixel values of corresponding teacher image pixel values to determine an error, and training the ANN to reduce the error. A method of detecting a target structure in an image involves scanning a local window across sub-regions of the image by moving the local window for each sub-region so as to obtain respective sub-region pixel sets, inputting the sub-region pixel sets to an ANN so that it provides respective output pixel values that represent likelihoods that respective image pixels are part of a target structure, the output pixel values collectively constituting a likelihood distribution map.

Abstract: Computerized aesthetic judgment of images is disclosed. In one embodiment, a computer-implemented method inputs a training set of images, where each image has a corresponding set of aesthetic scores. The method trains a classifier based on the training set, and outputs the classifier. An image can then be input into the classifier, such that an aesthetic score for the image is generated by the classifier and output. Furthermore, recommendations can be generated to improve the aesthetic score for the image, which are also output.

Abstract: A method, system, and computer program product of selecting a set of training images for a massive training artificial neural network (MTANN). The method comprises selecting the set of training images from a set of domain images; training the MTANN with the set of training images; applying a plurality of images from the set of domain images to the trained MTANN to obtain a corresponding plurality of scores; and determining the set of training images based on the plurality of images, the corresponding plurality of scores, and the set of training images. The method is useful for the reduction of false positives in computerized detection of abnormalities in medical images. In particular, the MTAAN can be used for the detection of lung nodules in low-dose CT (LDCT). The MTANN consists of a modified multilayer artificial neural network capable of operating on image data directly.

Abstract: An image processing device and method for classifying symbols, such as text, in a video stream employs a back propagation neural network (BPNN) whose feature space is derived from size, translation, and rotation invariant shape-dependent features. Various example feature spaces are discussed such as regular and invariant moments and an angle histogram derived from a Delaunay triangulation of a thinned, thresholded, symbol. Such feature spaces provide a good match to BPNN as a classifier because of the poor resolution of characters in video streams.

Abstract: A quadrature mirror filter is applied to decompose an image into at least two sub-images each having a different resolution. These decomposed sub-images pass through self-organizing map neural networks for performing a non-supervisory classification learning. In a testing stage, the recognition process is performed from sub-images having a lower resolution. If the image can not be identified in this low resolution, the possible candidates are further recognized in a higher level of resolution.

Abstract: A system permitting one or more users to navigate a wide-angle image, and in particular a panoramic image, using a neural network to correct distortion in that image. To train the neural network, a calibration pattern which defines an array of calibration points is disposed to occupy a field of view of a wide-angle imaging apparatus, and a calibration image is thereby captured by the apparatus. Respective view directions of the calibration points and the positions of these points in the calibration image are used as data for training the neural network to correctly match view directions to corresponding intra-image positions.

Abstract: A system and method are disclosed for determining the pose angle of an object in an input image. In a preferred embodiment, the present system comprises a pose estimator having a prototype projector, a regression estimator, and an angle calculator. The prototype projector is preferably adapted to reduce the input image dimensionality for faster further processing by projecting the input pixels of the image onto a Self-Organizing Map (SOM) neural network. The regression estimator is preferably implemented as a neural network and adapted to map the projections to a pattern unique to each pose. The angle calculator preferably includes a curve fitter and an error analyzer. The curve fitter is preferably adapted to estimate the pose angle from the mapping pattern. The error analyzer is preferably adapted to produce a confidence signal representing the likelihood of the input image being a face at the calculated pose.

Abstract: A new neural model for direct classification, DC, is introduced for acoustic/pictorial data compression. It is based on the Adaptive Resonance Theorem and Kohonen Self Organizing Feature Map neural models. In the adaptive training of the DC model, an input data file is vectorized into a domain of same size vector subunits. The result of the training (step 10 to 34) is to cluster the input vector domain into classes of similar subunits, and develop a center of mass called a centroid for each class to be stored in a codebook (CB) table. In the compression process, which is parallel to the training (step 33), for each input subunit, we obtain the index of the closest centroid in the CB. All indices and the CB will form the compressed file, CF. In the decompression phase (steps 42 to 52), for each index in the CF, a lookup process is performed into the CB to obtain the centroid representative of the original subunit. The obtained centroid is placed in the decompressed file.

Abstract: A new neural model for direct classification, DC, is introduced for acoustic/pictorial data compression. It is based on the Adaptive Resonance Theorem and Kohonen Self Organizing Feature Map neural models. In the adaptive training of the DC model, an input data file is vectorized into a domain of same size vector subunits. The result of the training (step 10 to 34) is to cluster the input vector domain into classes of similar subunits, and develop a center of mass called a centroid for each class to be stored in a codebook (CB) table. In the compression process, which is parallel to the training (step 33), for each input subunit, we obtain the index of the closest centroid in the CB. All indices and the CB will form the compressed file, CF. In the decompression phase (steps 42 to 52), for each index in the CF, a lookup process is performed into the CB to obtain the centroid representative of the original subunit. The obtained centroid is placed in the decompressed file.

Abstract: A divide-and-conquer (DAC) method and system improve the detection of abnormalities, like lung nodules, in radiological images via the use of zone-based digital image processing and artificial neural networks. The DAC method and system divide the lung zone into different zones in order to enhance the efficiency in detection. Different image enhancement techniques are used for each different zone to enhance nodule images, as are different zone-specific techniques for selecting suspected abnormalities, extracting image features corresponding to selected abnormalities, and classifying the abnormalities as either true or false abnormalities.