Abstract: A method and system to generate vectors that represent linearly progressing entities like time are disclosed. Traditional methods of vectorisation account for semantic or associative similarity of the entities. Thus, vectors conveying semantic information do not convey structural relations between such entities. The method allows for the representation of such structural information, for example the months in a year. The vectors generated by the invention encode this relation between the months such that one can interpret the sequence of the months, the difference between then and their cyclic nature. The method works in a manner similar to a genetic code, where subsequent “child” vectors are generated by related “parents”, thus encoding the similarity and the distance of the sequential entities. An object of the inventions to allow algorithms in machine learning to easily learn over temporal entities its natural text.

Abstract: A method of adaptive chroma downsampling is presented. The method comprises converting a source image to a converted image in an output color format, applying a plurality of downsample filters to the converted image and estimating a distortion for each filter chose the filter that produces the minimum distortion. The distortion estimation includes applying an upsample filter, and a pixel is output based on the chosen filter. Methods for closed loop conversions are also presented.

Abstract: An image processing apparatus includes: a first storing unit for storing a plurality of prediction parameters used in predictive image generation for a plurality of prediction units corresponding in size to at least one picture; a second storing unit for storing two or more prediction parameters used in predictive image generation for two or more prediction units among the plurality of prediction units; and a transferring unit configured to collectively transfer, for predictive image generation, (i) two or more of the prediction parameters from the first storing unit to the second storing unit or collectively transfer (ii) the two or more prediction parameters used in the predictive image generation for the two or more prediction units from the second storing unit to the first storing unit.

Abstract: Methods and devices for creating, compressing and searching a binary tree are provided by the present disclosure, which relates to the field of traffic classification. Binary tree compression includes: determining a number of a compression layer or a number of an intermediate node, compressing a binary tree according to the number of the compression layer or the number of the intermediate node to form at least one compression node, and creating a bitmap of the compression node. Therefore, the depth of the decision tree is reduced, the search speed is increased, rule expansion is avoided, and rule replication is reduced.

Abstract: Disclosed is a method for four-path tree structured vector quantization, comprising: (a) placing codebook vectors where magnitudes of vectors are arranged in order at lowermost nodes; (b) configuring upper nodes by averaging vector values of the lowermost adjacent four nodes; (c) calculating mean square errors to paths of the upper nodes (A, B, C, and D) corresponding to four paths from an input vector ‘X’; (d) presetting a threshold between 0 and 1 after the ‘(c)’; (e) calculating thresholds between the input vector ‘X’ and four paths ‘A’, ‘B’, ‘C’, and ‘D’, based on the mean square errors; and (f) comparing the thresholds calculated in the ‘(d)’ with the preset threshold to select a path having a threshold greater than the preset threshold, in which dynamic four-path tree structured vector quantization is used instead of conventional dynamic two-path tree structured vector quantization, thereby more efficiently and quickly searching a codebook.

Abstract: A 3D graphics rendering pipeline is used to carry out data comparisons for motion estimation in video data encoding. Video data for the pixel block of the video frame currently being encoded is loaded into the output buffers of the rendering pipeline. The video data for the comparison pixel blocks from the reference video frame is stored as texture map values in the texture cache of the rendering pipeline. Once the sets of pixel data for comparison have been stored, the rendering pipeline is controlled to render a primitive having fragment positions and texture coordinates corresponding to the data values that it is desired to compare. As each fragment is rendered, the stored and rendered fragment data is compared by fragment compare unit and the determined differences in the data values are accumulated in an error term register.

Abstract: A 3D graphics rendering pipeline is used to carry out data comparisons for motion estimation in video data encoding. Video data for the pixel block of the video frame currently being encoded is loaded into the output buffers of the rendering pipeline. The video data for the comparison pixel blocks from the reference video frame is stored as texture map values in the texture cache of the rendering pipeline. Once the sets of pixel data for comparison have been stored, the rendering pipeline is controlled to render a primitive having fragment positions and texture coordinates corresponding to the data values that it is desired to compare. As each fragment is rendered, the stored and rendered fragment data is compared by fragment compare unit and the determined differences in the data values are accumulated in an error term register.

Abstract: This disclosure describes adaptive filtering techniques to improve the quality of captured imagery, such as video or still images. In particular, this disclosure describes adaptive filtering techniques that filter each pixel as a function of a set of surrounding pixels. An adaptive image filter may compare image information associated with a pixel of interest to image information associated with a set of surrounding pixels by, for example, computing differences between the image formation associated with the pixel of interest and each of the surrounding pixels of the set. The computed differences can be used in a variety of ways to filter image information of the pixel of interest. In some embodiments, for example, the adaptive image filter may include both a low pass component and high pass component that adjust as a function of the computed differences.

Abstract: Exemplary embodiments of the present invention include a method for creating a user metric pattern. Such embodiments typically include receiving, within the network, a plurality of disparate user metrics, determining that the plurality of disparate user metrics received within the network do not match a predetermined metric pattern, and saving the plurality of disparate user metrics as a new metric pattern. In many embodiments, determining that the plurality of disparate user metrics received within the network do not match a predetermined metric pattern includes comparing the plurality of disparate user metrics with a plurality of metrics associated with the predetermined metric pattern.

Abstract: In an encoding method of moving pictures which generates a predictive picture for a current picture based on a reference picture and a motion vector, a macroblock is divided into subblocks. In each of the plurality of subblocks, an initial value of the motion vector is set and an evaluated value E on a difference between the current picture and the reference picture is calculated along a steepest descent direction to determine the minimum value. Then, the smallest evaluated value is selected among the minimum values obtained on the plurality of subblocks to determine the motion vector based on the pixel position of the smallest value.

Abstract: Every frame represented by a moving picture signal is divided into blocks. Calculation is made as to a number of pixels forming portions of a caption in each of the blocks. The calculated number of pixels is compared with a threshold value. When the calculated number of pixels is equal to or greater than the threshold value, it is decided that the related block is a caption-containing block. Detection is made as to a time interval related to the moving picture signal during which every frame represented by the moving picture signal has a caption-containing block. A 1-frame-corresponding segment of the moving picture signal is selected which represents a caption-added frame present in the detected time interval.