Seiko Epson Corporation

Google Inc.

Seiko Espoo Corporation

Viresh Ratnakar

Vasudev Bhaskaran

Frances Bordwell Haugen

Ashok Chhabedia Popat

Victor Ivashin

Onur Gonen Guleryuz

Chia-Hsin Li

Amitabh Singhal

Joseph Kieran O'Sullivan

Guillaume Raphael Poncin

Maxim Lifantsev

Amitabh K Singhal

Joseph Kieran O'sullivan

Text is extracted from a source image of a publication using an Optical Character Recognition (OCR) process. A document is generated containing text segments of the extracted text. The document includes a control module that responds to user interactions with the displayed document. Responsive to a user selection of a displayed text segment, a corresponding image segment from the source image containing the text is retrieved and rendered in place of the selected text segment. The user can select again to toggle the display back to the text segment. Each text segment can be tagged with a garbage score indicating its quality. If the garbage score of a text segment exceeds a threshold value, the corresponding image segment can be automatically displayed instead.

Selective display of OCR'ed text and corresponding images from publications on a client device

A system identifies a document relating to a search term, where the document includes a set of structural elements. The system determines a distribution of occurrences of the search term in the document, identifies one of the structural elements based on the distribution of occurrences of the search term in the document, and presents information associated with the identified structural element.

Presentation of search results based on document structure

Semantically meaningful segments of an image of a document, such as tables of contents, page numbers, footnotes, and the like, are identified. These segments form a model of the document image, which may then be rendered differently for different client devices. The rendering may be based on a display parameter provided by a client device, such as a display resolution of the client device, or a requested display format.

Display of document image optimized for reading

A layered presentation system (LAPE) includes a server that performs compressed-domain image processing on image data received from multiple clients including a master client and other clients to generate a composite image that incorporates imagery from the other clients with a master image from the master client for viewing on a shared display. The system's clients can add imagery in the form of questions, comments, and graphics to a currently displayed image. The added imagery is processed along with the master image to generate the composite image that then appears on the shared display and perhaps also on each client's individual display. The processing includes scaling the master image/added imagery, as required, and blending and/or overlaying the added imagery onto the master image so as to augment but not obscure it. A network protocol is included for sending image data in the compressed domain back and forth between the server and each of the clients.

LAPE: layered presentation system utilizing compressed-domain image processing

An efficient compressed-domain-based algorithm that works by combining the IDCT and pixel-domain averaging into a single compressed domain operation, where a part of the cost of even this simplified operation can be absorbed into the de-quantization process. The original image need not be decompressed fully, only to the point of extracting the transform coefficients. Compressed-domain scaling for all scales k/8 for 1&#x2266;k&#x2266;8, including &#x215c;, &#x215d;, &#xbe;, &#x215e;, is provided. By chaining multiple operations, any scale factors of the form kk. . . k/8for integers k's in the range [1,7] can be achieved.

Image scaling in the compressed domain

Visible watermarking techniques for digital images and video operate on decoded transform coefficients and maintain the watermarked data in the same compressed format as the original so that the watermarked data can be viewed with standard tools and applications. Moreover, for most of the techniques presented, the watermarked data has exactly the same compressed size as the original. The watermark can be inserted and removed using a key for applications requiring content protection. The watermark insertion and removal algorithms are very efficient and exploit some features of compressed data formats (such as JPEG and MPEG) which allow most of the work to be done in the compressed domain.

Standards compliant watermarking for access management

A compressed-domain-based algorithm reduces a source image to a given target size using a combination of downscaling, cropping, and region-of-interest identification. The source image is partitioned into a plurality of macroblocks, each macroblock containing a plurality of chrominance and luminance blocks, each chrominance block and each luminance block containing a DC coefficient and AC coefficients. To each macroblock, a first rule is applied based on values of the DC coefficient of the chrominance blocks in that macroblock to identify a particular type of texture in the image. Also to each macroblock, a second rule is applied based on select values of the AC coefficient of luminance blocks in that macroblock to identify edges in the image. The macroblocks within the image are then searched to find an area containing an area of interest based on the results of applying the first and second rules.

Semantic downscaling and cropping (SEDOC) of digital images

Image partitioning methods, which may be implemented by software or hardware on a computer, partition a set of multiple overlapping images to be transmitted so that only the visible portions of the images are transmitted. The underlying partitioning algorithm uses rectangles to iteratively partition the image set.

Image partitioning to avoid overlap transmission

A wavelet transform coding technique is provided, which may be part of TROBIC or may be used independently in a wide variety of compression applications. The wavelet coding technique is carried out by a simple block codec with wavelets (SBCW) that uses wavelet filters to transform 32&times;2 image blocks into subband coefficients organized in a tree structure. These coefficients are then encoded using a novel entropy coding technique with precise rate control. The Two Row Buffer Image Compression (TROBIC) technique operates on 32&times;2 blocks of pixels and uses a block-based coding scheme to provide alternative coding modes for image blocks containing different types of image data. The TROBIC algorithm automatically evaluates the image presented to determine regions containing text, graphics, synthetic images and natural images along with determining the most effective combination of global-index coding, local-index coding, lossless coding and wavelet coding to use to guarantee the desired compression ratio and high quality.

Wavelet transform coding technique

A steganographic method embeds hidden information like digital watermarks and digital fingerprints into an image by applying one or more zero-mean patches to a digital signal that conveys the image. Each zero-mean patch comprises elements having an average value substantially equal to zero. A printer tracking system is implemented by selecting a plurality of zero-mean patches according to a string of bits representing a printer serial number, and modifying printer control signals to embed the plurality of patches in each of a number of virtual cells within the image to be printed. The offset of the virtual cells with respect to the boundaries of the image and the location of the patches within each virtual cell are established in pseudo-random fashion. The printer tracking system further includes a technique for subsequently detecting the printer serial number by scanning the printed image and running a detection algorithm on the scanned data to recover the string of bits embedded in the zero-mean patches.

Watermarking with random zero-mean patches for printer tracking

A steganographic method embeds hidden information like digital watermarks and digital fingerprints into an image by applying one or more zero-mean patches to a digital signal that conveys the image. Each zero-mean patch comprises elements having an average value substantially equal to zero. A copyright protection system is implemented by selecting one or more zero-mean patches in pseudo-random fashion, modifying the patches by perturbing their elements in pseudo-random fashion, and applying the patches to image pixels at locations that are selected in pseudo-random fashion. The copyright protection system further includes a procedure for checking for the presence of a watermark in a digital signal.

Watermarking with random zero-mean patches for copyright protection

An original image is sharpened by obtaining a first frequency-domain representation of the original image, selecting one or more elements from this first representation based on one more criteria such as element magnitude and frequency, scaling the selected elements according to one or more scale factors, and forming a second frequency-domain representation by combining the scaled selected elements with the unselected elements of the first representation. A sharpened reproduction of the original image may be generated by applying an inverse transform to the second frequency-domain representation. A technique for deriving the value of the one or more scale factors is also discussed.

Growing community of inventors

Sunnyvale, CA, United States of America

Viresh Ratnakar