Abstract: Image templates are extracted from video images in real-time and stored in memory. Templates are selected on the basis of their ability to provide useful positional data and compared with regions of subsequent images to find the position giving the best match. From the position data a transform model is calculated. The transform model tracks the background motion in the current image to accurately determine the motion and attitude of the camera recording the current image. The transform model is confirmed by examining pre-defined image templates. Transform model data and camera sensor data are then used to insert images into the live video broadcast at the desired location in the correct perspective. Stored templates are periodically updated to purge those that no longer give valid or significant positional data. New templates extracted from recent images are used to replace the discarded templates.

Abstract: Image templates are extracted from video images in real-time and stored in memory. Templates are selected on the basis of their ability to provide useful positional data and compared with regions of subsequent images to find the position giving the best match. From the position data a transform model is calculated. The transform model tracks the background motion in the current image to accurately determine the motion and attitude of the camera recording the current image. The transform model is confirmed by examining pre-defined image templates. Transform model data and camera sensor data are then used to insert images into the live video broadcast at the desired location in the correct perspective. Stored templates are periodically updated to purge those that no longer give valid or significant positional data. New templates extracted from recent images are used to replace the discarded templates.

Abstract: Image templates are extracted from video images in real-time and stored in memory. Templates are selected on the basis of their ability to provide useful positional data and compared with regions of subsequent images to find the position giving the best match. From the position data a transform model is calculated. The transform model tracks the background motion in the current image to accurately determine the motion and attitude of the camera recording the current image. The transform model is confirmed by examining pre-defined image templates. Transform model data and camera sensor data are then used to insert images into the live video broadcast at the desired location in the correct perspective. Stored templates are periodically updated to purge those that no longer give valid or significant positional data. New templates extracted from recent images are used to replace the discarded templates.

Abstract: A method and apparatus for providing multiple interactive video teleconferencing sessions over a standard ISDN line is described. Multiple sets of audio, video, and data information are generated and transmitted to a bandwidth allocation device which determines what portion of that information will be transmitted over the ISDN line. Additionally, priority request values are generated that assist the bandwidth allocation device in performing its allocation function such that users interacting with the audio, video and data generation sources will experience acceptable levels of interaction even when bandwidth demand is greater than the supply.

Abstract: A computer-implemented method and apparatus for encoding image signals. According to a preferred embodiment, an image signal is received. A strength S of a variable low-pass filter having a variable strength is selected. The low-pass filter means is applied at strength S to the image signal. The filtered image signal is encoded to provide an encoded bitstream representative of the current image signal.

Abstract: In a digital video motion estimation compression and decompression system, pixel block-matching is accomplished by comparing pixels in target block regions of high gradients of luminance or hue. A few pixels from low-gradient regions are also preferably used. A mask defining these pixel locations in the target block is created, and the block comparisons with other blocks are based only on the relative values associated with the pixels in these locations. Major computational time savings are accomplished with negligible degradation of image quality.

Abstract: A system and method for realtime occlusion processing for seamlessly and realistically blending an inserted image such as an advertisement into a region of a live broadcast image without obscuring the action of the live image. The average color and intensity of a synthetic reference image containing at least some of the region to be replaced is compared to the average color and intensity of the current live broadcast image to determine the difference between the two images. The resulting difference image obtained from processing the current image and synthetic, reference image determines areas of the intended insertion region within the current image which are obscured by live action. The processor then generates an occlusion mask based on the difference image and only those pixels that are unoccluded within the intended insertion region are allowed to be inserted into the live broadcast.

Abstract: A computer-implemented method and apparatus for encoding video pictures of a sequence of video pictures, wherein each picture is defined by one picture type of a plurality of picture types and each picture has a complexity. According to a preferred embodiment, the complexity of each picture type is estimated to provide a complexity estimate for each picture type. Encoding a picture at an average quantization level produces an encoded bit stream having a codesize related to the complexity of the picture, and each picture type is associated with a target codesize. An average quantization level is determined for the first picture of each picture type in accordance with the target codesize and complexity estimate of each of the first pictures, respectively. The first picture of each picture type is then encoded at its respective average quantization level.

Abstract: An intelligent start for a block-matching search is employed. Motion vectors which have been calculated for immediately adjacent blocks in the same image and for corresponding blocks in previous images are used. If the region upon which motion estimation is being performed is uniform and there is no disturbance motion estimation may be very fast. If there are many discontinuities, for example edges, then a more extensive search is required. To do this, possible candidate vectors are searched and a comparison measure is obtained for each. From these the best match is determined . A local fine search is then performed. The best motion vector can usually be found by examining only a few well-chosen candidate vectors. This greatly improves the efficiency of the search.

Abstract: A computer-implemented method and apparatus for encoding pictures of a sequence of pictures. According to a preferred embodiment, an encoding system determines a relatively fixed background of the sequence of pictures, and transmits the background to a decoding system. Foreground objects of a current picture are separated from the background, and motion compensation encoding of the foreground objects is performed with respect to at least one previous picture. The encoded foreground objects are then transmitted to the decoding system.

Abstract: A computer-implemented method and apparatus for processing signals. According to a preferred embodiment, a rate controller determines an initial quantization level for quantizing a block of coefficient signals representative of an image, wherein coefficient signals quantized with a given quantization level are associated with a quantization error. At least one new quantization level is generated in accordance with the initial quantization level. The quantization error associated with the initial quantization level and the at least one new quantization level is determined. A quantization level is selected from the initial quantization level and the at least one new quantization level in accordance with this determination.

Abstract: A computer-implemented method for decoding motion-compensated blocks of a picture of a sequence of pictures. According to a preferred embodiment, a current motion-compensation encoded block is decoded. Embedded block edges within the current block are determined, and a low-pass filter is applied along the embedded block edges.

Abstract: A computer-implemented process and apparatus are used for efficient conversion of color video signals. The most significant bits (MSBs) of YUV data are masked off and converted into a single index that allows for the number of colors needed in a standard display format. The Y MSBs are scaled down by a fractional factor prior to the formation of the index so as to save computer memory space that would otherwise be wasted. In a preferred embodiment, relatively low-noise and temporally-changing dithering is applied to the values of full-sampled YUV data prior to color conversion.

Abstract: A computer-implemented method and apparatus for encoding a current block of pixels of a current picture of a sequence of pictures, the sequence of pictures comprising a previous picture having a previous reference block and a subsequent picture having a subsequent reference block. A previous residual error is determined between the current block and the previous reference block, and a subsequent residual error is determined between the current block and the previous reference block. At least one interpolated picture interpolated between the previous and subsequent pictures is defined, the interpolated picture having an interpolated reference block. An interpolated residual error is determined between the current block and the at least one interpolated reference block from the previous and subsequent residual errors.

Abstract: A computer-implemented method and apparatus for processing video signals representative of pixels of a current block of a current picture. According to a preferred embodiment, the current block is compared to a plurality of blocks of a previous picture to provide a plurality of corresponding motion vectors, wherein each motion vector corresponds to a respective motion estimation error measure. It is determined whether the current block contains at least one region of high spatial energy and at least one region of relatively low spatial energy associated with different motion vectors of the plurality of corresponding motion vectors. The block is encoded in accordance with this determination.

Abstract: A computer-implemented method and apparatus for upsampling subsampled chroma pixels of a picture. According to a preferred embodiment, there is provided a plurality of subsampled chroma pixels as well as a plurality of luma pixels. The subsampled chroma pixels are upsampled as a function of luma pixels of the plurality of luma pixels to provide at least one upsampled chroma pixel.

Abstract: A computer-implemented method and apparatus for encoding video signals, wherein the encoding is defined by a plurality of encoding parameters. According to a preferred embodiment, an initial quantization level is determined for quantizing a block of signals representative of a block of pixels of a picture, wherein the initial quantization level is an encoding parameter of the plurality of encoding parameters. At least one encoding parameter of the plurality of encoding parameters is modified before encoding the block of signals to provide a modified plurality of encoding parameters such that block edge artifacts between the block of pixels and at least one adjacent block of pixels that arise after decoding encoded signals are reduced.

Abstract: Variable-length encoded signals (e.g., those corresponding to video signals) are decoded using a lookup table whose indices are shorter than the longest possible variable-length encoded signal. N bits from the encoded bit stream are used as the index to retrieve from the lookup table a table entry, which comprises a flag bit indicating whether the variable-length encoded signal is a short code signal or a long code signal, one or more code-length bits, and one or more code-value bits. If the flag bit indicates that the variable-length encoded signal is a short code signal, then the decoded signal is generated using the code-value bits, wherein the code-length bits indicate the length of the variable-length encoded signal.

Abstract: Encoded video signals comprise sets of transform coefficients (e.g., DCT coefficients) corresponding to different regions of a video frame. An inverse transform is applied to sets of transform coefficients to generate decoded regions of a decoded video frame. The discontinuities for boundaries between adjacent regions are used to adjust one or more of the transform coefficients. The adjusted sets of transform coefficients are then used to generate filtered regions of a filtered video frame corresponding to the decoded video frame. In a preferred embodiment, the transform coefficients are DCT coefficients and the DC and first two AC DCT coefficients are sequentially adjusted to correct for quantization errors in the encoding process.

Abstract: Video frames to be encoded are divided into regions, which are then classified into different classes of regions. One or more different filters are then applied to the regions of different classes to generate pre-filtered video frames which are then further encoded to generate an encoded bit stream. In a preferred embodiment, the regions are classified into substantially changed regions and substantially unchanged regions. The pre-filtered video frames are generated by applying a temporal filter to the substantially unchanged regions and by applying a spatial filter to the substantially changed regions.