Abstract: A motion estimator has a spatial sub-sampler to receive input images; at least one motion estimator determining motion vectors between input images and sub-sampled motion vectors between sub-sampled images; an up-sampler for up-sampling the sub-sampled motion vectors; and a selector for providing a motion vector output by selecting between the motion vectors and the (up-sampled) sub-sampled motion vectors, according to motion vector confidence.

Abstract: A method of identifying the left-eye and the right-eye images of a stereoscopic pair, comprising the steps of comparing the images to locate an occluded region visible in only one of the images; detecting image edges; and identifying a right-eye image where image edges are aligned with a left hand edge of an occluded region and identifying a left-eye image where more image edges are aligned with a right hand edge of an occluded region.

Abstract: To assess picture impairment due to the interpolation of output images from input images, for example in standards conversion, an output image detail measure is compared with an input image detail measure. One of the image detail measures is obtained by interpolation between image detail measures for at least two images.

Abstract: A method of image segmentation using graph-theory techniques, in which pixels of an image are represented by the vertices of a graph. A minimum spanning tree is generated and edges of the tree are successively removed according to a morphological property to leave a spanning forest, trees of which correspond to segments of the image. Selection of edges for removal may depend on an energy function of a tree and on the energy function of trees created by removal of that edge.

Abstract: For monitoring an image transformation such as aspect ratio conversion, an image feature is defined by identifying a position in the image having a local spatial maximum value and then identifying four other positions in the image having local spatial minimum values such that the four minimum value positions surround the position of the maximum, a first pair of the minimums lie on a first line passing through maximum and a second pair of the minimums lie on a second line passing through the maximum.

Abstract: Systems and methods to detect non-uniform spatial scaling of an image in the horizontal direction (for example in 4:3 to 16:9 aspect ratio conversion).

Abstract: To modify the aspect ratio of images in a video sequence, a measure of image energy is provided at or around each pixel, for example by taking differences between adjacent pixels. A minimum energy path of contiguous pixels is then identified which extends across the image from one edge to another, for example vertically if the width of the picture is to be modified. Information concerning the location of a path in one image is used in the identification of a path in the next image, for example by biassing the energy measure of the path. The aspect ratio of each image is then modified by removing or by replicating the minimum energy path of contiguous pixels.

Abstract: Video data is segmented by representing the pixel location, RGB values and other features such as motion vectors, as points in a multidimensional segmentation space. Initialized segments are represented as locations in the segmentation space and segment membership then determined by the distance in segmentation space from the data point representing the pixel to the location of the segment. The distance measure takes into consideration the covariance of the data, for the segment or for the picture.

Abstract: In image processing, parallel and synchronous pixel processing elements have a pixel value register, a neighbor value register, and a processor receiving adjacent pixel and neighbor values for four adjacent pixels. In a series of iterations, the neighbor value register is updated with—in one mode—the maximum and—in another mode—the minimum of: the current value; the pixel value from any adjacent pixel whose pixel value is different from the current pixel value and the neighbor value from an adjacent pixel whose pixel value is the same as the current pixel value. The pixel value is then replaced by the minimum or maximum (depending on the mode) of the current pixel and neighbor values. The modes may alternate with the number of iterations in each series remaining constant or growing.

Abstract: To enable efficient use of limited bandwidth in transmitting video, a region of interest is determined in each image. Before coding, the image is spatially scaled, with magnification applied inside that a region of interest. The scaled images are then compression encoded. Meta-data identifying the location of the region of interest accompanies the transmitted video so that, after decoding, the scaling can be reversed.

Abstract: A video program length is changed by processing an MPEG compressed video signal. Two temporally adjacent B-frames are selected, by looking for pairs with small differences. The selected B-frames are combined to create at least one intermediate frame; which is inserted in place of or between the selected frames. Presentation time stamps are altered to provide a uniform display rate.

Abstract: The invention concerns image processing and, in particular, the processing of picture attribute fields for an image. A method of obtaining a new picture attribute field of an image is disclosed in which a picture attribute value at one position is allocated to a new position in the image in dependence upon the value of a parameter, such as luminance data, at the original position and at the new position and/or in dependence on the distance between the original position and the new position. The invention may be used to process picture attribute fields comprising: motion vectors; motion vector confidence; segment labels; depth labels; texture labels.

Abstract: Systems and methods for detecting a border region in an image. A blank border in a video picture is determined by summing luminance or other pixel measures in a direction parallel to the border and looking for the maximum gradient of those summed measures in a direction perpendicular to the border. Sensitivity can be enhanced by increasing (relative to other pixels) the gain of pixels around the present pixel value of the border. The location of the maximum gradient may be weighted by other measures before a decision on border location is taken.

Abstract: A moving image sequence is automatically reframed for a small display. The framing is adjusted in dependence upon foreground and background segment weights derived for pixels in images in the sequence. Reframed images are formed from only those input pixels which fall within a reframing window within the input image area and the position or size of the reframing window is adjusted so as to maximise a weighted total of the sum of the foreground weights of pixels within the window and the sum of the background weights of pixels outside the window.

Abstract: In the creation of video program material, picture composition decisions—especially framing parameters—taken by an operator for small-display presentation of a scene are used the automatic generation of picture composition decisions for a larger display. Framing parameters such as pan, are temporally filtered before being applied to the wider field of view provided for the larger display.

Abstract: In a video encoder a prediction signal at a bit resolution of P is subtracted from an input signal received at a bit resolution of P+?, where P is typically 8 bits, and P+?, 10 bits. A transform is then performed on the result, the output of which transform being at a bit resolution of at least P+?. The transformed signal is quantised, and the quantised signal used to form the prediction signal. The quantised signal is then coded for downstream delivery. At a decoder, the process is inverted, preserving the extra ? bits input to the encoder.

Abstract: In a video encoder a prediction signal at a bit resolution of P is subtracted from an input signal received at a bit resolution of P+?, where P is typically 8 bits, and P+?, 10 bits. A transform is then performed on the result, the output of which transform being at a bit resolution of at least P+?. The transformed signal is quantized, and the quantized signal used to form the prediction signal. The quantized signal is then coded for downstream delivery. At a decoder, the process is inverted, preserving the extra ? bits input to the encoder.

Abstract: A video signal is analyzed to determine what compression coding decisions would be taken if the video signal were to be MPEG2 encoded. A representation of these coding decisions then accompanies the un-coded video signal for use in a downstream “dumb” encoding process. The coding decisions take the form of an MPEG2 bitstream and are embedded in Isb's of the video.

Abstract: This invention concerns the spatial interpolation of colour images and, in particular, the reconstruction, or “de-mosaicing” of data from a single sensor-array electronic camera. Electronic cameras typically have an image sensor comprising a matrix of individual pixel sensors, each sensor being responsive to a colour component. In order to obtain colour component information for all pixels, in accordance with the disclosed embodiment, green component pixel values (203) are interpolated to obtain green component pixel values for all pixels (209). A difference value (B?G) (212) is formed from the green colour component values (207) and the original blue component values (204) at blue pixel locations, and the difference values (B?G) (212) are then interpolated to obtain difference values for all pixels (216). Blue component values for all pixels (225) can then be obtained from the difference values for all pixels (216) and the green component values for all pixels (209).

Abstract: In compression coding of picture sequences, a measure of picture difference is generated and compared with a picture difference threshold. The comparison is used in selecting between the use of predicted (P-pictures) and non-predicted has an initial value adapted to promote the use of an I-picture at or close to a scene change in the picture sequence, and changes in time in such a manner as to increase the probability of selection of a I-picture with an increase in the time since selection of the last I-picture.