Daewoo Electronics Co., Ltd.

Min-Sup Lee

In a three-dimensional model-based coding system, a shape of mouth, especially, a shape of lip, among a shape of face, has an important visual information. As each person has the different shape of lip, however, it is desirable to transform the shape of lip based on an individual person. First, a characteristic three-dimensional lip model based on the individual person is generated by matching a three-dimensional lip model to a basic lip of the individual person. Subsequently, the characteristic three-dimensional lip model is converted into a two-dimensional lip model and rotated according an angle of a actual lip image. In case that the person is talking, the two-dimensional lip model is reconstructed based on a recognized syllable. Finally, a difference between the two-dimensional lip model and the actual lip image is calculated. As not a whole real lip image but the difference between the actual image and the two-dimensional lip model similar to the actual lip image is transmitted, the amount of transmission data can be reduced.

Method and apparatus for producing lip-movement parameters in a

An apparatus for predicting a current frame signal included in an image signal based on a reference frame first detects a panning vector representing a spatial displacement between the current and the reference frames. Once the panning vector is detected, a predicted frame containing a substitutable region and an unsubstitutable region is determined based on the panning vector and the reference frame, and then an amplitude of the panning vector is compared with a reference value. Subsequently, a predicted current frame signal is generated as the result of the current frame prediction process in response to the result of the comparison. If the amplitude of the panning vector is smaller than the reference value, the predicted current frame signal is produced by padding pixel values of the substitutable region which are located along a boundary between the substitutable region and the unsubstitutable region into the unsubstitutable region. Meanwhile, if the amplitude of the panning vector is greater than or equal to the reference value, the predicted current frame signal is created by using a block matching technique for the unsubstitutable region based on the current and the reference frame signals.

Current frame prediction method and apparatus for use in an image signal

A video signal coding system employing a segmentation technique segments a video frame signal into regions to generate contour and texture information of each region and detects continuous boundaries located between adjacent regions. The contour and the texture information of each region and the boundary information of the continuous boundary are encoded to provide an encoded video signal. The video signal coding system calculates a mean value of pixels in two adjacent regions of each side of a continuous boundary and updates each pixel value within a predetermined distance from a continuous boundary with the mean value.

Video signal coding system employing segmentation technique

An apparatus, for use in a low bit-rate image signal encoder, for detecting a foreground region in an image signal, wherein the image signal has a number of objects therein and the foreground region has at least one moving object therein, which comprises: a previous frame memory for storing a previous frame of the image signal; a current frame memory for storing a current frame of the image signal; a change detector for comparing the current frame and the previous frame to detect changed pixels, each changed pixel having a differential luminance value; a region detector for detecting an object region for one of the objects in the current frame; and a foreground/background determinator for counting the number of all of the pixels contained in the object region to determine a first size of the object region; and for counting the number of changed pixels which are located in the object region to determine a second size of the object region; and for comparing the first size with the second size to determine whether the object region is the foreground region.

Apparatus for detecting a foreground region for use in a low bit-rate

An apparatus for determining a set of motion vectors for a set of feature points located on a contour of a moving object of a video signal comprises a contour displacement detection circuit for detecting a contour displacement range between a previous contour of the moving object contained in the previous frame and a current contour of the moving object contained in the current frame; and a feature point motion estimation circuit for comparing one of feature point blocks, having one of the feature points, with each of candidate blocks derived from a corresponding feature point search region adaptively regulated by the contour displacement range and for assigning a spatial displacements between one of the feature point block and a most similar candidate block thereto as a motion vector for a feature point corresponding to the one to thereby generating the set of motion vectors for the set of feature points.

Apparatus for estimating motion vectors for feature points of a video

An apparatus for tracing contours of segmented regions in an image includes a zero masking unit to produce a closed contour loop for each of the segmented regions, a tracing circuit for sequentially assigning each of the non-zero masked pixels on the closed contour loop as a tracing pixel, comparing the tracing pixel with a set of its adjacent pixels located at a rightward, an upward, a leftward and a downward directions with respect to the tracing pixel to select one of the adjacent pixels having a luminance level identical to that of the tracing pixel and updating the tracing pixel with the selected adjacent pixel with until the updated tracing pixel coincides with the tracing pixel selected first.

Apparatus for tracing contours of segmented regions

A motion-compensated video signal encoder has a circuit for determining a predicted current frame based on a current frame and a previous frame of a digital video signal. The circuit includes a region detection circuit for detecting a processing region encompassing a moving object from the previous frame based on a difference between the current and the previous frames to generate region information representing the detected processing region. Therefore, a number of pixels is selected from the pixels contained in the detected processing region as feature points based on the region information. A first set of motion vectors between the current and the previous frames, each of the first set of motion vectors representing a motion for each of the selected pixels is then detected. The first set of motion vectors is used for predicting the predicted current frame and transmitted as a set of motion vectors of the video signal together with the region information.

Apparatus for encoding a video signal using feature point based motion

An apparatus for encoding a contour of regions contained in a video signal comprises first and second segmentation blocks for segmenting previous and current frame into a number of regions to generate previous and current contours, a prediction parameter calculation block for generating prediction parameters based on the previous and current contours, and the previous and current frames, a contour prediction block for constructing a predicted current contour by transforming the previous contour based on the prediction parameters, an error detection block for detecting deviation points lying on portions of the predicted current contour that deviate from the current contour, thereby generating the deviation points and corresponding displacements to the deviation points, and a contour coder for encoding the deviation points and the displacements together with the prediction parameters.

Apparatus for encoding a contour of regions contained in a video signal

A method for determining a set of motion vectors between a current frame and a reference frame of video signals, which comprises the steps of: (a) one dimensionally comparing a search block from the current frame with said plurality of candidate blocks included in a search region corresponding to the search block, on a block-by-block basis, by employing a one-dimensional error function using horizontal and vertical integral projections to select a predetermined number of candidate blocks in an ascending order of the one-dimensional error; (b) two dimensionally comparing the search block with the predetermined number of candidate blocks selected in step(a) above, on a block-by-block basis, by employing a two-dimensional error function to thereby select a most similar candidate block and derive a motion vector representing the displacement of pixels between the search block and the most similar candidate block so as to assign the derived motion vector as the motion vector for the search block; and (c) repeating steps (a) and (b) until all of the search blocks constituting the current frame are searched against the entire reference frame, to thereby derive a set of motion vectors.

Method for detecting motion vectors

A method, for use in a segmentation-based coding system, for detecting motion vectors between a current frame and a previous frame of digital video signal comprises the steps of (a) generating segmentation data for each of the segmented regions, the segmentation data including contour information representing the shape and location for each segmented region and texture information representing a mean gray level of all pixels contained in each segmented region, (b) generating candidate segmentation data for each of the candidate regions, (c) producing a minimum-sized predetermined geometric pattern for encompassing a segmented region, (d) producing a minimum-sized predetermined geometric pattern for encompassing each of the candidate regions, (e) calculating an error value between the segmented region and each of the candidate regions, (f) selecting a candidate region yielding a smallest error value as a search region, (g) detecting one or more motion vectors between the segmented region and the search region; and (h) repeating said steps (c) to (g) for all the segmented regions.

Method for detecting motion vectors for use in a segmentation-based

A method for providing an accurately predicted current frame, comprising the steps of: selecting feature points in the previous frame and detecting a first set of motion vectors for the feature points; defining quadrangles by connecting the feature points; defining horizontally split triangles by splitting quadrangles and calculating a variance for each triangle and comparing the variance to select the smallest variance as first variance; defining vertically split triangles by splitting the quadrangles and calculating the variance for each triangle and comparing the variance to select the smallest variance value as second variance value; comparing the first and the second variance and selecting the smaller variance as a third variance and providing triangle information representing the split direction for the third variance; determining quasi-feature points on the current frame and a second set of motion vectors for the quasi-feature points; defining quadrangles formed by connecting the quasi-feature points, and splitting each quadrangle into two triangles; determining a predicted position on the previous frame for pixel contained in triangle of the current frame; determining a third set of motion vectors for the pixels contained in triangle of the current frame; and providing value of each pixel to be contained in the predicted current frame by using the second and the third set of motion vectors.

Method for encoding a video signal using feature point based motion

An apparatus, for use in an image signal encoding system, encodes a bidirectionally predictive frame (B frame) of a digital video signal to generate an encoded B frame signal. The apparatus selects two groups of pixels from pixels contained in a detected processing regions of a first and a second reference frames as a first and a second sets of feature points, respectively, by using a feature points selection circuit; and then determines a first set of motion vectors between the current and the first reference frames and a second set of motion vectors between the current and the second reference frames through the use of a feature point motion estimation circuit. Thereafter, the apparatus selects a set of motion vectors for the set of feature points from the first and the second sets of motion vectors to thereby generate a selected set of motion vectors and selection information representing the selected set as the encoded B frame.

Method and apparatus for encoding and decoding a video signal using

An inventive apparatus estimates correct motion vectors between a current and its previous frames for feature points selected from the previous frame. A motion vector detector first determines a set of initial motion vectors for the feature points. Subsequently, a displacement generator provides N number of displacements to an initial motion vector of a selected feature point to thereby produce a set of N candidate motion vectors for the selected feature point, wherein the selected feature point is surrounded by its, e.g., six neighboring feature points to form a previous hexagon having a set of six triangles. And then, a predictor provides quasi-feature points in the current frame from the neighboring feature points through the use of the initial motion vectors thereof; and also generates a set of N selected quasi-feature points from the selected feature point by using the N candidate motion vectors. And a weight factor generator selectively generates one from a multiplicity of weigh factors based on position data of each of the pixels in the current hexagon and each of the N subject quasi-feature points. Finally, a motion vector selector selects a motion vector from the N candidate motion vectors corresponding to a set of six triangles in the current hexagon which yields a minimum error value among N error values obtained through the use of the weight factors between the current and the previous hexagons.

Motion vector estimation method and apparatus for use in an image signal

A method for detecting and thinning a contour image of an object in a video frame comprises the steps of: (a) calculating edge magnitudes for pixels in the video frame and detecting edge points, (b) generating a contour image constituted by the detected edge points, and (c) thinning the contour image by comparing an edge magnitude for each of the edge points with each of the edge magnitudes included in a plurality of edge magnitude sets, each of the edge magnitude sets having a series of three edge magnitudes selected from the edge magnitudes corresponding to the neighboring pixels surrounding said each of the edge points.

Method and apparatus for detecting and thinning a contour image of

An apparatus, for use in a motion-compensated video signal encoder, for determining a predicted current frame based on a current and a previous frame, comprises: a feature point motion vector(MV) detector for detecting a first set of MVs for feature points; an edge point MV detector for, in each of the edges, detecting one or two neighboring feature points for each of edge points, and determining a second set of MVs for the edge points; a quasi-point detector for detecting quasi-feature points and a third set of MVs therefor, and detecting quasi-edge points and a fourth set of MVs therefor; a non-quasi-point MV detector for detecting a fifth set of MVs for all non-quasi-points, each of the fifth set of MVs being determined by averaging the MVs of one or more neighboring quasi-feature points for each of the non-quasi-points; an arrangement block for arranging the third, the fourth and the fifth sets of MVs to determine a group of MVs for all of the pixels in the current frame; and a motion compensator for providing a pixel value from the previous frame based on the position data of each pixel in the current frame and a MV thereof.

Growing community of inventors

Seoul, South Korea

Min-Sup Lee