Abstract: In one implementation, we improve the binary voxel-based octree coding method, via a proposed state summarization module for context modeling. Given a current voxel to be encoded or decoded, instead of directly estimating its occupancy probability based on the associated binary occupancy context, a proposed state summarization module is applied to convert the original binary context to a summarized representation. Under the summarized representation, the estimation of the occupancy probability becomes more affordable and effective. In particular, density-based state summarization, pattern-based, learning-based state summarization, and learning-based state summarization methods are provided.

Abstract: A Folding-based point cloud compression system or a regular convolution-based point cloud compression system may lack the ability to precisely manage compression of all points in a point cloud frame. For example, they may provide good predicted positions for some points but not necessarily an exact reconstruction, especially after being quantized. Consequently, some outlier points may fall outside of the surface constructed by a Folding-based method or a convolution-based method. In one implementation, we propose to compress two types of outlier points separate from “inlier” points. First, we detect model outlier points, which are points that do not have good prediction from a model. Then quantization outlier point detection is conducted to further separate those points that are sensitive to the compression (e.g., quantization) in the model transformed domain. Such outlier processing would provide efficient coding of most points (inlier points) and special handling on the outlier points.

Abstract: In one implementation, a learnable transformation TearingTransform over 3D point cloud data is proposed. The TearingTransform could decompose point clouds into two channels: a low rank channel and a sparse channel. The low rank channel corresponds to a codeword representing a rough shape of a point cloud. The sparse channel appears as an image-like data representing residual information that can refine the reconstructed point locations. In an encoder based on TearingTransform, a PN module is used to generate the codeword from the input point cloud; a FN module is used to reconstruct a preliminary point cloud from the codeword and an initial grid image; and a TN module modifies the initial grid image to generate an adjusted grid image. The codeword and the adjusted grid image are compressed. At the decoder, the point cloud can be reconstructed based on the decompressed codeword and adjusted grid image.

Abstract: In one implementation, we improve the binary voxel-based octree coding method, via a proposed state summarization module for context modeling. Given a current voxel to be encoded or decoded, instead of directly estimating its occupancy probability based on the associated binary occupancy context, a proposed state summarization module is applied to convert the original binary context to a summarized representation. Under the summarized representation, the estimation of the occupancy probability becomes more affordable and effective. In particular, density-based state summarization, pattern-based, learning-based state summarization, and learning-based state summarization methods are provided.

Abstract: In one implementation, we propose the UnfoldingOperator, which unfolds/flattens an unorganized input 3D point cloud onto a regular 2D grid. Given an input point cloud, an input 2D grid and the reconstructed point cloud produced by the FoldingNet, our proposal maps the input point cloud onto the 2D grid based on the reconstructed point cloud, leading to a 3-channel image. Alternatively, instead of using an image alone to represent a point cloud, the point cloud is decomposed into a codeword and a 3-channel residual image. This residual image is obtained by subtracting the reconstructed point cloud from the original input. The proposed UnfoldingOperator can be applied to point cloud compression, leading to a corresponding compression system that we call UnfoldingCompression. The UnfoldingCompression can work with the TearingCompression, where we can adaptively choose whether to use the UnfoldingCompression or TearingCompression.

Abstract: In one implementation, a learnable transformation TearingTransform over 3D point cloud data is proposed. The TearingTransform could decompose point clouds into two channels: a low rank channel and a sparse channel. The low rank channel corresponds to a codeword representing a rough shape of a point cloud. The sparse channel appears as an image-like data representing residual information that can refine the reconstructed point locations. In an encoder based on TearingTransform, a PN module is used to generate the codeword from the input point cloud; a FN module is used to reconstruct a preliminary point cloud from the codeword and an initial grid image; and a TN module modifies the initial grid image to generate an adjusted grid image. The codeword and the adjusted grid image are compressed. At the decoder, the point cloud can be reconstructed based on the decompressed codeword and adjusted grid image.

Abstract: Method of transcoding video data with fusion of coding units, computer program, transcoding module and telecommunications equipment associated therewith. Method of transcoding video data between a first and a second format (F1, F2), the method comprising a step of decoding the binary stream (FB1) providing decoded video data, data representative of the coding structure of the frames in the first format (F1) and, for all or some of the first coding units, prediction data, and a step of re-encoding in the course of which the decoded video data are encoded in the second format (F2). During the re-encoding step, an intermediate coding structure is constructed, comprising intermediate coding units constructed so as to correspond to the fusion of one or more first coding units, prediction data are allocated to each of the intermediate coding units, and the decoded video data are re-encoded in the second format (F2) as a function of the intermediate coding structure.

Abstract: A key frame representative of a sequence of frames in a video file is selected. The sequence of frames is divided into shots, sequences of frames captured by a single continuous operation of the camera, by detecting shot boundaries. Shot boundaries are detected by measuring changes in motion activity between frames using measures of pixel intensity and histogram differences between frames. The most interesting shot in the sequence of frames is selected using measures of motion activity, spatial activity, skin pixel color and shot length. The key frame is selected in the most interesting shot using measures of spatial activity and motion activity.

Abstract: This invention describes a video surveillance system which is composed of three key components 1—smart camera(s), 2—server(s), 3—client(s), connected through IP-networks in wired or wireless configurations. The system has been designed so as to protect the privacy of people and goods under surveillance. Smart cameras are based on JPEG 2000 compression where an analysis module allows for efficient use of security tools for the purpose of scrambling, and event detection. The analysis is also used in order to provide a better quality in regions of the interest in the scene. Compressed video streams leaving the camera(s) are scrambled and signed for the purpose of privacy and data integrity verification using JPSEC compliant methods. The same bit stream is also protected based on JPWL compliant methods for robustness to transmission errors.

Abstract: A camera-based document scanning system produces electronic versions of documents, based on a plurality of images of discrete portions of the documents. The system compares each pair of consecutive images and derives motion parameters that indicate the relative motion between each pair of consecutive images. The system utilizes the derived motion parameters to align and merge each image with respect to the previous images, thereby building a single, mosaic image of the document. In the illustrative embodiment, the motion parameters are derived by minimizing a sum of squared differences equation on a pixel-by-pixel basis.

Abstract: A key frame representative of a sequence of frames in a video file is selected. The sequence of frames is divided into shots, sequences of frames captured by a single continuous operation of the camera, by detecting shot boundaries. Shot boundaries are detected by measuring changes in motion activity between frames using measures of pixel intensity and histogram differences between frames. The most interesting shot in the sequence of frames is selected using measures of motion activity, spatial activity, skin pixel color and shot length. The key frame is selected in the most interesting shot using measures of spatial activity and motion activity.

Abstract: A key frame representative of a sequence of frames in a video file is selected. The sequence of frames is divided into shots, sequences of frames captured by a single continuous operation of the camera, by detecting shot boundaries. Shot boundaries are detected by measuring changes in motion activity between frames using measures of pixel intensity and histogram differences between frames. The most interesting shot in the sequence of frames is selected using measures of motion activity, spatial activity, skin pixel color and shot length. The key frame is selected in the most interesting shot using measures of spatial activity and motion activity.

Abstract: A key frame representative of a sequence of frames in a video file is selected by applying face detection to a video to select a key frame which may include people and has particular application to indexing video files located by a search engine web crawler. A key frame, one frame representative of a video file, is extracted from the sequence of frames. The sequence of frames may include multiple scenes or shots, for example, continuous motions relative to a camera separated by transitions, cuts, fades and dissolves. To extract a key frame face detection is performed in each frame and a key frame is selected from the sequence of frames based on a sum of detected faces in the frame.

Abstract: A technique for organizing data information in a network having a plurality of network stations is disclosed. In one embodiment, the technique is realized by having a first processing device store a representation of data at an address of a first of the plurality of network stations. A second processing device then stores the address at a second of the plurality of network stations in association with an identifier of the data. A third processing device then stores the data identifier at a third of the plurality of network stations in association with an annotation of the data.

Abstract: A camera-based document scanning system produces electronic versions of documents, based on a plurality of images of discrete portions of the documents. The system compares each pair of consecutive images and derives motion parameters that indicate the relative motion between each pair of consecutive images. The system utilizes the derived motion parameters to align and merge each image with respect to the previous images, thereby building a single, mosaic image of the document. In the illustrative embodiment, the motion parameters are derived by minimizing a sum of squared differences equation on a pixel-by-pixel basis.

Abstract: A method and apparatus is described for encoding a video sequence of frames. Each frame in the video sequence is organized in blocks of pixels. A scene change is detected when a current frame in the video sequence is substantially different from a previous frame. When it is determined that the current frame is the change in scene, the current frame is coded to be an intra frame with each block of pixels of the intra frame is being an intra-coded block. Coding the sequence of frames produces a compressed bit stream having a coded intra frame at each scene change. Each coded intra frame provides an access point in the bit stream from which a storyboard of the scenes in the video sequence can be generated.

Abstract: In an coder for producing a bitstream representative of a sequence of video images, a previous image is registered with a current image using spline-based registration to produce estimated motion vectors. The estimated motion vectors are used to match blocks of the previous image and the current image to produce translation vectors. The translation vectors compensate for motion while encoding the sequence as a bitstream.

Abstract: A technique for processing data is disclosed. In one embodiment, the technique is realized by receiving a first representation of data at a processing device, which then processes the first representation of data so as to generate a second representation of data. The second representation of data includes a plurality of dependent data representations and a plurality of independent data representations. Each of the plurality of dependent data representations is substantially aligned in time with a corresponding one of the plurality of independent data representations.

Abstract: A method and apparatus is described for encoding a sequence of video frames at a target bit rate. A controller controls the bit r ate by providing values for a coding frame rate and quantization parameter to a frame encoder. A set of operating regions, including a first operating region and a second operating region, is defined. Each operating region includes a range of values for each parameter. These operating regions may intersect each other or be disjoint. The encoder codes the frame sequence with the value of each parameter being in the first operating region. During the coding, the controller determines to make an adjustment to the value of one parameter that would put that value out of the first operating region and in the second operating region. The controller makes the adjustment if a predetermined criterion is satisfied, otherwise the controller constrains the value of that one parameter to remain in the first operating region.

Abstract: A technique for accessing an item of interest within a particular one of a plurality of stored representations of data is disclosed. In one embodiment, the technique is realized by having a processing device searching a plurality of stored annotations corresponding to different items within the plurality of stored representations to locate an annotation of interest corresponding to the item of interest. The annotation of interest has an associated search identifier and an associated location identifier corresponding to a location of interest within the particular one of the plurality of stored representations. The processing device then searches a plurality of stored search identifiers associated with the plurality of stored annotations to locate the search identifier and an address identifier corresponding to a location of the particular one of the plurality of stored representations within the plurality of stored representations.