Abstract: A pool of messages, e.g., e-mails and/or other electronic documents that each correspond to a communication from a sender to a recipient, is analyzed to identify communication chains between a source and a target. Sender and recipient identifiers extracted from the messages are used to detect direct and indirect communication links between pairs of entities. Information related to the identified communication chains can be presented to a user via an interactive network graph that supports iterative analysis of the communication-chain data.

Abstract: Reliable identification of highly similar documents allows such documents to be treated as identical for purposes of document analysis. Identification of highly similar documents can be based on a composite hash value or other value for which the likelihood of two documents having the same value is high if and only if the documents have a high degree of similarity. Prior to performing content based analysis, the composite hash value for the current document is determined and compared to composite hash values of previously analyzed documents. If a match is found, the results of the analysis of the previous document can be applied to the current document. If no match is found, the current document is analyzed.

Abstract: Reliable identification of highly similar documents allows such documents to be treated as identical for purposes of document analysis. Identification of highly similar documents can be based on a composite hash value or other value for which the likelihood of two documents having the same value is high if and only if the documents have a high degree of similarity. Prior to performing content based analysis, the composite hash value for the current document is determined and compared to composite hash values of previously analyzed documents. If a match is found, the results of the analysis of the previous document can be applied to the current document. If no match is found, the current document is analyzed.

Abstract: A pool of messages, e.g., e-mails and/or other electronic documents that each correspond to a communication from a sender to a recipient, is analyzed to identify communication chains between a source and a target. Sender and recipient identifiers extracted from the messages are used to detect communication links between pairs of entities. Indirect chains of any desired length can be found by iteratively tracing a communication path one step forward from the source, then one step backward from the target, and so on; at each new step, entities at end points of the forward paths and backward paths are compared to detect any entities that complete a communication chain from source to target. Information related to the identified communication chains can be presented to a user via an interactive report that supports iterative analysis of the communication-chain data.

Abstract: A photo extraction technique is provided that scans film contained within a holder designed to accommodate that particular size and type of film and generates a low-resolution scan of the film and holder. The low-resolution scan is then processed by detecting the borders of the holder, applying a smoothen filter, detecting edge segments of the image-holding areas in the holder, identifying each of the image-holding areas, determining the orientation of the image-holding areas and correcting the orientation thereof if necessary, and locating each of the individual photos within the image-holding areas. Having identified and extracted the photos, an index page comprising thumbnail images of the photos is generated. From this page, the user can select one or more of the images to be printed or displayed in high-resolution.

Abstract: Improved data embedding and extracting techniques provide a way to embed and extract messages in text sections of documents during copying. Extracted text pixels are grouped together to form text lines of the document. From this formation, a document layout is constructed that is used to embed the message in the text pixels. Each text line is partitioned into blocks, and those of which contain a certain threshold percentage of text pixels are identified as valid. Each valid block is used to embed one bit of information by labeling text pixels of that block with a certain predetermined color. The embedding of bits in valid blocks in a particular text line is done in a column-wise raster order. Only one message character (which may be comprised of multiple bits) is embedded in a particular text line, although that character may be embedded multiple times in the same line if there are enough valid blocks.

Abstract: Adaptive filtering techniques for hardware or software implementation, which selectively use median and low-pass filters to reduce artifacts based on local document features, are disclosed. Specific techniques for low- and high-resolution documents are provided. A fast and memory efficient algorithm for computing the median on a 3×3 pixel neighborhood is also disclosed, and is used in the filtering technique for high-resolution documents.

Abstract: We improve text reproduction of a scanned and printed image by identifying text-contribution colors and also image-contribution colors that are in or near the text region of the color space. When input colors are encountered that contribute both to text and image we splice together text and image color transformations of the input color to produce a color in the printer color space. We improve background reproduction, i.e. reduce bleed-through artifacts, by identifying background-contribution colors and also image-contribution colors that are in or near the background region of the color space. When input colors are encountered that contribute both to background and image we splice together background and image color transformations of the input color to produce a color in the printer color space.

Abstract: A photo extraction technique is provided that scans film contained within a holder designed to accommodate that particular size and type of film and generates a low-resolution scan of the film and holder. The low-resolution scan is then processed by detecting the borders of the holder, applying a smoothen filter, detecting edge segments of the image-holding areas in the holder, identifying each of the image-holding areas, determining the orientation of the image-holding areas and correcting the orientation thereof if necessary, and locating each of the individual photos within the image-holding areas. Having identified and extracted the photos, an index page comprising thumbnail images of the photos is generated. From this page, the user can select one or more of the images to be printed or displayed in high-resolution.

Abstract: A photocopier (10) scans a document and so modifies the resultant image as to embed in a resultant copy information that is visually imperceptible. To do so, it identifies regions (32, 33) that consist primarily of text-character parts. It embeds a message bitstream by selectively darkening the text pixels in such regions in accordance with the values of successive bits in the bitstream.

Abstract: A color image reproduction system achieves higher-quality color reproduction by improving the utilization of colors within the gamut of an output device that are not in the gamut of an input device. This is accomplished by a device-dependent compensation transformation that maps a second set of colors in both the gamut of an input device and the gamut of the output device into a first set of colors in the gamut of the output device but not in the gamut of the input device. The compensation transformation may be derived in a number of ways that entail identifying the first and second sets of colors and then determining one or more scaling factors that map the second set of colors into a union of the first and second sets of colors.

Abstract: An image processor (14) in a photocopier (10) takes composite input image (20a) that includes a plurality of skewed sub-images (22a, 24a, and 26a). It revises the image so as to produce an output composite image (20b) containing aligned sub-images (22b, 24b, and 26b). To identify the input image's sub-images, the image processor (14) identifies connected groups (44, 48 and 54) of pixels that meet certain criteria, and it eliminates from consideration each group (48) whose bounding box (50) is included within the bounding box (46) of another connected group (44). Skew angles for the different sub-images are determined by finding the best-fit lines through the outline pixels of their respective connected groups, and the sub-images are de-skewed by the negatives of those lines' angles.

Abstract: A scanner (12) derives a digital image of a document that it scans optically, an image-processing circuitry (14) extracts a representation of a non-planner graph. A small-processor “smart card” derives from the first graph a second graph that is isomorphic to it and related to it in accordance with a secret permutation matrix. The image processor 14 then embeds a representation of that graph into the image and sends the results to a printer (16) to generate a copy. A second scanner (30) generates a digital image of the copy, and processing circuitry (31) extracts the two isomorphic graphs, which it conveys to a verifier circuit (36) as well as the smart card (22). By repeatedly generating and submitting to the verifier (36) test graphs that are isomorphic to the extracted graphs, the smart card (22) can demonstrate, without revealing the secret permutation matrix, that it is in possession of that permutation matrix.

Abstract: A scanner (12) of a photocopier system scans a document and generates output image signals that an image-processing circuit (14) processes in such a manner as to enhance text-region legibility, and a laser printer (16) generates the output copy in accordance with the resultant image. The image-processing circuitry (14) identifies dark, low-saturation pixels in the input image whose gradients are low but that border high-gradient pixels, and it sets these pixels to the blackest values, thus emphasizing text edges. With the remaining pixels, it reduces bright-region contrast, and this tends to suppress the visibility of text that has “bled through” from the document's reverse side or from documents disposed behind it.

Abstract: Dither thresholds are assigned one after the other to matrix locations in the process of generating a dither matrix used for printer half-toning. The matrix location to be assigned the next threshold is chosen by locating the tightest cluster or largest void in the dot pattern that will result from the gray level with which the threshold being assigned is associated. Measures of cluster tightness for low-range and high-range thresholds are based on the areas of Voronoi partitions associated with respective candidate locations. For mid-range thresholds, a Gaussian-filter output is used as the measure. In both cases, ties between candidate locations are resolved by applying a further criterion, which depends on the candidate locations' proximities to locations assigned thresholds the same as the one being assigned or differing from it by only one.

Abstract: A system for detecting image black and white points for a digital image.

Abstract: An improved display of a color image is achieved by an optimum placement of color dots using an efficient stochastic screening process in which display elements are formed according to a priority ranking of color and according to a spatial-priority matrix. Various combinations of basic colors are ranked in priority according to visibility such that elements in a stochastic screen are first formed for the most visible colors. The number of elements in the screen that are used to represent more than one basic color is minimized. In preferred embodiments, a spatial-priority matrix is used to control the placement of colors in a stochastic screen such that an optimally uniform distribution of colors is achieved. A process for generating a spatial priority matrix for display devices that do not have a one-to-one display aspect ratio is described.

Abstract: A printer driver operates a printer capable of multiple dot-placement geometries or resolutions and multiple inks per color channel. It maps a continuous-tone cyan pixel-component value to continuous-tone light- and dark-cyan values, which it separately halftones to respective binary pixel values that respectively indicate whether an associated pixel will receive light- and dark-cyan ink dots. It similarly generates binary values that respectively indicate whether an associated pixel will receive light- and dark-magenta ink dots. In performing the halftoning, the driver separately performs error diffusion through different interleaved sub-images of the image to be printed so that the error-diffusion process is performed for images whose pixel geometry is that for which the process was designed even though the overall image's pixel geometry is not.

Abstract: An improved display of a color image is achieved by an optimum placement of color dots using an efficient stochastic screening process in which display elements are formed according to a priority ranking of color and according to a spatial-priority matrix. Various combinations of basic colors are ranked in priority according to visibility such that elements in a stochastic screen are first formed for the most visible colors. The number of elements in the screen that are used to represent more than one basic color is minimized. In preferred embodiments, a spatial-priority matrix is used to control the placement of colors in a stochastic screen such that an optimally uniform distribution of colors is achieved. A process for generating a spatial priority matrix for display devices that do not have a one-to-one display aspect ratio is described.

Abstract: A method and apparatus for segmenting a document which has both text and image regions. The method and apparatus implement a technique in which large text pixels and image pixels are identified in a document having a relatively low resolution. The method and apparatus then detect dark text pixels on a light background region of a document and assign segmentation labels to each pixel. The pixel labels are post-processed using a plurality of syntactic rules to correct mislabeling of pixels. This process does not change the visual perception of the image regions in the document. Pixels identified as being in the background region of the document are assigned a white label and pixels identified as being in the text region are assigned a black label. The resulting processed document contains sharp black text and white background, resulting in improved perceptual quality and efficient ink utilization during a printing process.