Abstract: A frequency-domain audio coder selects among different entropy coding modes according to characteristics of an input stream. In particular, the input stream is partitioned into frequency ranges according to some statistical criteria derived from a statistical analysis of typical or actual input to be encoded. Each range is assigned an entropy encoder optimized to encode that range's type of data. During encoding and decoding, a mode selector applies the correct entropy method to the different frequency ranges. Partition boundaries can be decided in advance, allowing the decoder to implicitly know which decoding method to apply to encoded data. Or, adaptive arrangements may be used, in which boundaries are flagged in the output stream by indicating a change in encoding mode for subsequent data. For example, one can create a partition boundary which separates out primarily zero quantized frequency coefficients, from primarily non-zero quantized coefficients, and then apply a coder optimized for such data.

Abstract: A technique for entropy coding information relating to frequency domain audio coefficients. For portions of a frequency spectrum having a predominate value of zero, a multi-level run length encoder statistically correlates sequences of zero values with non-zero values and assigns variable length code words. An encoder uses a specialized code book generated with respect to the probability of receiving an input sequence of zero-valued spectral coefficients followed by a non-zero coefficient. A corresponding decoder associates a variable length code word with a sequence of zero value coefficients adjacent a non-zero value coefficient.

Abstract: A human speech detection method detects pure-speech signals in an audio signal containing a mixture of pure-speech and non-speech or mixed-speech signals. The method accurately detects the pure-speech signals by computing a novel Valley Percentage feature from the audio signal and then classifying the audio signals into pure-speech and non-speech (or mixed-speech) classifications. The Valley Percentage is a measurement of the low energy parts of the audio signal (the valley) in comparison to the high energy parts of the audio signal (the mountain). To classify the audio signal, the method performs a threshold decision on the value of the Valley Percentage. Using a binary mask, a high Valley Percentage is classified as pure-speech and a low Valley Percentage is classified as non-speech (or mixed-speech). The method further employs morphological filters to improve the accuracy of human speech detection.

Abstract: A sprite generation method used in video coding generates a sprite from the video objects in the frames of a video sequence. The method estimates the motion between a video object in a current frame and a sprite constructed from video objects for previous frames. Specifically, the method computes motion coefficients of a 2D transform that minimizes the intensity errors between pixels in the video object and corresponding pixels inside the sprite. The method uses the motion coefficients from the previous frame as a starting point to minimizing the intensity errors. After estimating the motion parameters for an object in the current frame, the method transforms the video object to the coordinate system of the sprite. The method blends the warped pixels of the video object with the pixels at corresponding positions in the sprite using rounding average such that each video object in the video sequence provides substantially the same contribution to the sprite.

Abstract: A segmentation method segments selected objects from other objects within a video image frame. The objects may be of arbitrary configuration and preferably represent distinct image features in a display image. An interior outline is formed within the interior of the object perimeter. The interior outline is expanded automatically to form an exterior outline. Pixels between the interior outline and the exterior outline are classified according to predefined attributes as to whether they are within object interior, thereby to identify automatically a conformed outline that is fitted to the object perimeter.

Abstract: To represent the motion of a pixel between successive image frames, this method utilizes the motion information for the transformation block containing the pixel, as well as the motion information for adjacent transformation blocks whenever the adjacent transformation blocks are associated with the same object with which the selected pixel is associated. As a result, transformation errors arising from transformation block discontinuities are decreased and the accuracy and integrity of motion representations of pixels within an object are maintained. Preferably, the motion representations for the transformation blocks are multi-dimensional or affine transformations that are capable of representing complex motions that include any or all of translation, rotation, magnification, and shear.

Abstract: A transformation method provides a multi-dimensional affine transformation for representing motion between corresponding image components of successive video image frames. The multi-dimensional affine transformations can represent complex motion that includes any or all of translation, rotation, magnification, and shear. The transformation method of this invention includes determining motion transformations between corresponding pixels in the image components of the first and second video image frames. From the motion transformations between corresponding pixels, multi-dimensional affine motion transformations between the corresponding image components are determined. This transformation method increases the accuracy with which complex motion is represented and results in fewer compression or encoding errors in comparison to conventional methods, thereby increasing compression efficiency.

Abstract: A video compression encoding process compresses or encodes dense motion vector fields to provide and improved compression ratio. In accordance with this method, a dense motion vector field is obtained for an object or a portion of an object between a pair of video image frames. The dense motion vector field may correspond, for example, to an object or other image portion of arbitrary configuration or size. The configuration of the dense motion vector field is extrapolated to a regular, preferably rectangular, configuration to facilitate encoding or compression. The dense motion vector field with its extrapolated regular configuration is encoded or compressed according to conventional encoding transformations such as, for example, discrete cosine transformation (DCT) or lattice wavelet compression.

Abstract: Video coding efficiency for high motion scenes is improved by adaptively disabling a parameter indicating whether texture and motion data is coded for a macroblock. The COD parameter is disabled when the number of macroblocks with substantially all zero motion and texture data is less than a threshold number. This reduces the number of bits required to code an interframe video image in a video sequence with high motion and large changes from frame to frame. The coded block pattern for chrominance is also used to determine how to perform entropy coding for the coded block pattern for luminance. In interframe blocks, if the chrominance blocks are coded, it is likely that the luminance blocks will be coded as well. The coded block pattern for chrominance, therefore, is used to select the appropriate entropy coding table for the coded block pattern for luminance.

Abstract: A hierarchical object encoding technique or process capable of representing general binary arbitrary shapes that include, for example, embedded or disconnected components. The method decomposes successive layers of general binary arbitrary shapes into simple arbitrary shapes. Each mask formed in this manner is a simple arbitrary shape having only a continuous outer boundary. Accordingly, each outer boundary is encoded, preferably by a contour encoding method, to provide accurate encoding of general binary shapes.

Abstract: Video encoding and decoding processes provide compression and decompression of digitized video signals representing display motion in video sequences of multiple image frames. The encoder process utilizes object- or feature-based video compression to improve the accuracy and versatility of encoding interframe motion and intraframe image features. Video information is compressed relative to objects or features of arbitrary configurations, rather than fixed, regular arrays of pixels as in conventional video compression methods. This reduces the error components and thereby improves the compression efficiency and accuracy. The decoder process decompresses the encoded video information to reconstruct the objects or features of arbitrary configurations.

Abstract: A transformation block optimization method provides transformation blocks of variable size for video compression processes. The method includes defining a transformation block of multiple pixels in an image frame and determining a peak signal-to-noise ratio for the transformation block. The transformation block is tentatively subdivided into subcomponents for which a signal-to-noise ratio is also obtained. The subcomponents are adopted as transformation blocks whenever the signal-to-noise ratios of the subcomponents are greater than the signal-to-noise ratio of the transformation block by more than a predetermined threshold.

Abstract: Contraction and expansion of video images are important capabilities for multi-media, television picture-in-picture functionality, digital video archiving, browsing, and video transmission. A video image expansion process is used to expand contracted images. The video image expansion process uses a non-linear median filter to interpolate the original pixel values of the contracted image. The non-linear median filter provides good approximations of the original pixel values including pixel values in high contrast regions, such as boundary regions. The video image expansion process is fast, and provides excellent results for low bitrate video coding used for contracting and expanding video images. The video compression image expansion scheme is used with video compression encoding schemes such as MPEG to produce high quality expanded video images from contracted video images.

Abstract: A polygon block matching method includes defining a preliminary quadrilateral reference pixel block of multiple pixels about a selected reference pixel in an image object of a video image frame. The pixels of the preliminary quadrilateral pixel block not in the interior of the image object are identified and omitted to establish a reference pixel block that conforms to the image object. If it encompasses the perimeter of the image object, the reference pixel block frequently is of a non-quadrilateral polygonal configuration. A search or sample pixel block of multiple sample pixels is defined in another video image frame and represents a region over which a selected sample pixel corresponding to the selected reference pixel is sought. The search includes determining and storing for the pixels in the sample pixel block correlations to the pixels in the reference pixel block, and identifying from the correlations the selected sample pixel corresponding to the selected reference pixel.

Abstract: A video compression error signal in a video compression scheme is affected by random and high frequency impulse noise. An error signal suppressor containing two filters is applied to the video compression error signal. The first filter reduces or eliminates random noise. The second filter eliminates high frequency impulse noise. Random and high frequency noise is reduced or eliminated from frequencies that are unimportant to human visual perception. The error signal suppressor reduces the overall video compression bitrate by up to between 10% and 20% which provides corresponding increases in video compression and transmission efficiency. The error signal suppressor is used in video compression encoding schemes such as MPEG to reduce random and high frequency noise.

Abstract: A method of correlating transformation blocks between video frames for a block matching motion estimation process utilizes redundancy inherent in correlating pixel blocks to significantly reduce the processing resources required to perform the correlation. The method includes defining for the transformation blocks subsets of multiple pixels. In video frames that include pixels arranged as rows and columns of pixels, for example, the subsets can include columns of pixels in the transformation blocks. Correlations are obtained and stored for the transformation block subsets. Many of the correlations for the transformation block subsets of a transformation block match the correlations for the transformation block subsets of adjacent transformation blocks. Such matching correlations are retrieved rather than being recalculated as in conventional correlation methods.

Abstract: Sprite-defined objects are completely defined throughout the video sequence as of their first appearance by a "sprite" and one or more trajectories. The sprite includes all the image characteristics of an object throughout the video sequence, and one or more trajectories warp or transform the sprite to represent the object in each frame of the video sequence. The sprite-defined object or objects are a subset of the general objects in the general video sequence and have available more information when they first appear in a video sequence than general objects. A simplified compression method allows the additional information available for sprite-defined objects to be utilized more efficiently than the additional information would be by encoder and decoder processes for general objects.

Abstract: A hierarchical object encoding technique or process capable of representing general binary arbitrary shapes that include, for example, embedded or disconnected components. The method decomposes successive layers of general binary arbitrary shapes into simple arbitrary shapes. Each mask formed in this manner is a simple arbitrary shape having only a continuous outer boundary. Accordingly, each outer boundary is encoded, preferably by a contour encoding method, to provide accurate encoding of general binary shapes.

Abstract: A method implemented in an object-based video encoder or decoder uses shape information that describes the boundary of a group of pixels representing an object in a sequence of video frames to identify transparent blocks (e.g., macroblocks or blocks so that coding/decoding of these blocks can be skipped. In the object-based video coding method, encoders code shape separately from motion and texture, and shape information is available before the encoder/decoder codes/decodes texture and motion data. The encoder and decoder use this shape information to identify transparent macroblocks or blocks so that texture coding and possible motion coding can be skipped. This method for transparent block skipping reduces coding and decoding operations and reduces the number of bits needed to store a bitstream representing a compressed video sequence.

Abstract: Motion estimation plays a very important role in a video compression scheme. It is well known that in a video compression scheme, there exists a high temporal correlation between consecutive video image frames. However, it is difficult to do motion estimation for newly exposed objects, since there is no temporal information and thus, the error signal generated by newly exposed object is large. A method of unrestricted motion estimation is used to take advantage of spatial correlation between the pixels of a newly exposed object. Using unrestricted motion estimation, a newly exposed reference object is padded and expanded with virtual pixels which have the same characteristics of the portion of the object which is newly exposed. The padded object is used to determine the error signal. The error signals produced by using unrestricted motion estimation are significantly smaller, reducing the error signal by up to between 5% and 10%.