Hewlett-packard Development Company, L.p.

University of California

Hewlett-packard Company

Other

Algotec Systems Ltd.

Michael Elad

Doron Shaked

Peyman Milanfar

Sina Farsiu

Renato Keshet

Irwin E Sobel

Daniel Keren

Ronny Kimmel

Craig Gotsman

Oded Shmueli

Ron Kimmel

Carl Staelin

Darryl Greig

Ran Bar-Sella

Yacov Hel-Or

Renato Kresch

Danny Barash

Menashe Benjamin

Michael D Robinson

Yosef Reichman

Jacob Margolin

A method of creating a super-resolved color image from multiple lower-resolution color images is provided by combining a data fidelity penalty term, a spatial luminance penalty term, a spatial chrominance penalty term, and an inter-color dependencies penalty term to create an overall cost function. The data fidelity penalty term is an L1 norm penalty term to enforce similarities between raw data and a high-resolution image estimate, the spatial luminance penalty term is to encourage sharp edges in a luminance component to the high-resolution image, the spatial chrominance penalty term is to encourage smoothness in a chrominance component of the high-resolution image, and the inter-color dependencies penalty term is to encourage homogeneity of an edge location and orientation in different color bands. A steepest descent optimization is applied to the overall cost function for minimization by applying a derivative to each color band while the other color bands constant.

System and method for robust multi-frame demosaicing and color super resolution

A computer method of creating a super-resolved grayscale image from lower-resolution images using an Lnorm data fidelity penalty term to enforce similarities between low and a high-resolution image estimates is provided. A spatial penalty term encourages sharp edges in the high-resolution image, the data fidelity penalty term is applied to space invariant point spread function, translational, affine, projective and dense motion models including fusing the lower-resolution images, to estimate a blurred higher-resolution image and then a deblurred image. The data fidelity penalty term uses the Lnorm in a likelihood fidelity term for motion estimation errors. The spatial penalty term uses bilateral-TV regularization with an image having horizontal and vertical pixel-shift terms, and a scalar weight between 0 and 1. The penalty terms create an overall cost function having steepest descent optimization applied for minimization. Direct image operator effects replace matrices for speed and efficiency.

Robust reconstruction of high resolution grayscale images from a sequence of low resolution frames

An integrated method for both super-resolution and multi-frame demosaicing includes an image fusion followed by simultaneous deblurring and interpolation. For the case of color super-resolution, the first step involves application of recursive image fusion separately on the three different color layers. The second step is based on minimizing a maximum a posteriori (MAP) cost function. In one embodiment, the MAP cost function is composed of several terms: a data fidelity penalty term that penalizes dissimilarity between the raw data and the super-resolved estimate, a luminance penalty term that favors sharp edges in the luminance component of the image, a chrominance penalty term that favors low spatial frequency changes in the chrominance component of the image, and an orientation penalty term that favors similar edge orientations across the color channels. The method is also applicable to color super-resolution (without demosaicing), where the low-quality input images are already demosaiced. In addition, for translational motion, the method may be used in a very fast image fusion algorithm to facilitate the implementation of dynamic, multi-input/multi-output color super-resolution/demosaicing.

System and method for robust multi-frame demosaicing and color super-resolution

A method is provided of solving the dynamic super-resolution (SR) problem of reconstructing a high-quality set of monochromatic or color superresolved images from low-quality monochromatic, color, or mosaiced frames. The invention includes a joint method for simultaneous SR, deblurring, and demosaicing, this way taking into account practical color measurements encountered in video sequences. For the case of translational motion and common space-invariant blur, the proposed invention is based on a very fast and memory efficient approximation of the Kalman filter (KF). Experimental results on both simulated and real data are supplied, demonstrating the invention algorithms, and their strength.

Dynamic reconstruction of high-resolution video from color-filtered low-resolution video-to-video super-resolution

An interactive system for obtaining data for diagnostic purposes from a server having access to stores of said data wherein the server supplies the data and software to a data requesting user that enables the user to receive the data progressively so as to decide during the receipt of the data what portions of the data are to be received thereby decreasing the time required to receive the data and that enables the user to load the software to process the supplied data.

Data distribution system

A system and a method solve the estimation problem of finding reflectance R and illumination L. The system and method to solve a functional of the unknown illumination L such that a minimum of the functional is assumed to yield a good estimate of the illumination L. Having a good estimate of the illumination L implies a good estimate of the reflectance R. The functional uses a variational framework to express requirements for the optimal solution. The requirements include: 1) that the illumination L is spatially smooth; 2) that the reflectance values are in the interval [0,1] &#x2014;thus, when decomposing the image S, the solution should satisfy the constraint L&#x3e;S; 3) that among all possible solutions, the estimate of the illumination L should be as close as possible to the image S, so that the contrast of the obtained R is maximal; and 4) that the reflectance R complies with typical natural image behavior (e.g., the reflectance is piece-wise smooth).

System and method for image enhancement, dynamic range compensation and illumination correction

A method and apparatus for processing an image are disclosed. For edge detection, the zero crossing pixel location, n, within the image is identified. Then, a sub-pixel position offset, a, is determined. The offset is a relationship between interpolated image and an interpolated kernel at the zero crossing pixel location. Finally, the interpolated position for the edge location, t, is determined from the pixel location, n, and the sub-pixel position offset, a. For matching the detected edges of an object to edges of a model object, translation [v, v] between a detected edge of an object in the image and the corresponding edge of a model object is first determined. Next, cross correlation between the detected edge and the corresponding model edge is calculated. Then, translation error [e, e] is determined. Finally, the translation is updated.

Method and apparatus for sub-pixel edge detection

A method an apparatus uses a variational approach for space dependent gamut mapping. The method presents a new measure for the problem. A solution to the formulation according to the method is unique if the gamut of the target device is known. A quadratic programming efficient numerical solution provides real-time results.

Space varying gamut mapping

Demosaicing convolution kernels are incorporated into the framework of bilateral filtering in order to reduce artifacts at abrupt intensity transitions when a color mosaic pattern is converted to an output image. As a consequence of the bilateral filtering within a window that is selected without intensity considerations, intensity values of pixels that are physically close are given greater weight than intensity values of more distant pixels and, simultaneously, intensity values that are quantitatively similar (i.e., photometrically similar) are given greater weight than intensity values that are quantitatively dissimilar. Using photometric similarity in a demosaicing operation reduces the effects of pixels on one side of an abrupt intensity transition in determining interpolated intensity values for pixels on the opposite side of the abrupt intensity transition.

Bilateral filtering in a demosaicing process

This disclosure provides a system for classifying images, used in image detection, image recognition, or other computer vision. The system processes directory images to obtain eigenvectors and eigenvalues, and selects a set of &ldquo;smooth&rdquo; basis vectors formed by linear combinations of these eigenvectors to be applied against a target image. Contrary to conventional wisdom, however, a group of the eigenvectors having the weakest eigenvalues are used to select the basis vectors. A second process is then performed on this group of &ldquo;weakest&rdquo; eigenvectors to identify a set of candidate vectors, ordered in terms of &ldquo;smoothness.&rdquo; The set of basis vectors (preferably 3-9) is then chosen from the candidate vectors in order of smoothness, which are then applied in an image detection or image recognition process. Unlike some conventional systems where &ldquo;strong&rdquo; directory presence and thresholds are used to detect possible matches, the present system uses smooth, weak vectors to ideally produce zero or near zero results for matches.

Template matching system for images

Method and system for detecting and classifying objects in an image

A method of quickly and automatically comparing a new document to a large number of previously seen documents and identifying the document type. First, provide a plurality of document type distributions, each document type distribution describes layout characteristics of an independent document type and may include a plurality of data points. Each document type distribution includes data derived from at least one basis document signature which may include data defining pixels of a low-resolution image of the independent basis document resolved to between 1 and 75 dots per inch or may include document segmentation data derived from the independent basis document. Next provide a new electronic document. Then create new document signature from the new electronic document. Next, distances between the new document signature and each of the plurality of document type distributions are calculated using an algorithm based on a Bayesian framework for a Gaussian distribution. The distances calculated may be Euclidean distances or may be Mahalanobis distances. Additionally, calculating the distances may include weighting the value given each of a plurality of data points in the document signatures based on the usefulness of each of the plurality of data points in distinguishing between the document signatures. Next, select at least one candidate document type for the new electronic document from among the independent document types described by the plurality of document type distributions. The selection of the at least one candidate document type may include selecting a preselected fixed number of the independent document types or may include selecting the independent document types described by those of the plurality of document type distributions having calculated distances that are within a preselected threshold distance of the smallest of the distances calculated. In addition, the invention provides for a program storage medium readable by computer, tangibly embodying a program of instructions executable by the computer to perform the method steps described above.

Automatic categorization of documents using document signatures

A system and a method for classifying input vectors into one of two classes, a target class and a non-target class, utilize iterative rejection stages to first label the input vectors that belong in the non-target class in order to identify the remaining non-labeled input vectors as belonging in the target class. The system and method may be used in a number applications, such as face detection, where the members of the two classes can be represented in a vector form. The operation of the system can be divided into an off-line (training) procedure and an on-line (actual classification) procedure. During the off-line procedure, projection vectors and their corresponding threshold values that will be used during the on-line procedure are computed using a training set of sample non-targets and sample targets. Each projection vector facilitates identification of a large portion of the sample non-targets as belonging in the non-target class for a given set of sample targets and sample non-targets. During the on-line procedure, an input vector is successively projected onto each computed projection vector and compared with a pair of corresponding threshold values to determine whether the input vector is a non-target. If the input vector is not determined to be a non-target during the successive projection and thresholding, the input vector is classified as a target.

Growing community of inventors

Haifa, Israel

Michael Elad