Abstract: In segmenting multidimensional image data, the data is represented as a set of points in a vector space which is the product of the vector space of pixel values and the vector space of pixel locations. Segments are determined by the result of clustering of the points, typically by simulating gravitational clustering of the points considered as point masses. From time to time the points are merged according to a proximity criterion, typically defined by mutual occupancy of a pre-defined hypercube in the vector space. The mapping of original pixels to points is tracked during the simulation process, so that pixels may be mapped to segments.

Abstract: A de-interlacing process takes a weighted sum of pixels in a filter aperture to generate a pixel in an output picture, the weighted sum including products of triplets of pixels. Using a training sequence of progressive material, it is possible to calculate the weighting coefficients necessary to minimize the mean square error between the filter output and the desired result.

Abstract: In video scratch repair, scratches are treated as additive pedestals which are constant or slowly changing along the scratch. For a given picture, these pedestals are estimated and subtracted from the video input to reveal original picture detail. The pedestal can be estimated by heavily filtering along the direction of the presumed scratch, the output of a prior art scratch detector.

Abstract: A method for converting between relatively close field rates of perhaps 60 Hz and 59.94 Hz, monitors the phase difference between input and output field rates. Initially, output fields are created by synchronizing input fields. As the measured phase passes a threshold, the method switches to interpolation mode for a short, fixed interval and then returns to synchronization. The method offers much of the sharpness associated with simple synchronization but avoids the visible discontinuities of field drops or repeats.

Abstract: In video signal processing, diagonal lines or slopes are detected in the input material, and an adaptive spatial filter having positive, negative and linear filter apertures is employed. The positive filter aperture is employed upon detection of any positive slopes in excess of a defined positive threshold; the negative filter upon detection of negative slopes, and the linear filter otherwise.

Abstract: In a video noise reducer, where pixel information is recursively filtered, with recursion being disabled at a disjuncture (caused by motion, planning, shot-change or otherwise), recursively filtered pixel information prior to disjuncture continues to be output until pixel information following the disjuncture has been recursively filtered to an appropriate degree of noise reduction.

Abstract: A video signal is enhanced using a non-linear combination of left and right difference signals (L and R), constrained to be zero if either L or R is zero, using a minimum absolute value function. Edge enhancement is provided when the polarities of L and R are the same and peak enhancement when the polarities are different. Difference signals are derived with delays (2 and 4) and a subtracter (3). The non-liner combinations are derived in a lookup table (30).

Abstract: A method of display unconversion wherein each output field line is interpolated from at least three input field lines from each of at least three input fields. By increasing the field rate by 1.5 (instead of simple field doubling) and by increasing the line rate by 1.5 (instead of simple line doubling) a substantial increase in picture quality can be achieved while maintaining data flow rates within manageable limits.

Abstract: A method for converting between relatively dose field rates of perhaps 60 Hz and 59.94 Hz, monitors the phase difference between input and output field rates. Initially, output fields are created by synchronising input fields. As the measured phase passes a threshold, the method switches to interpolation mode for a short, fixed interval and then returns to synchronization. The method offers much of the sharpness associated with simple synchronization but avoids the visible discontinuities of field drops or repeats.

Abstract: A group delay of a noise-reducing median filter for video or a video FIR filter is reduced by transferring filter contributions of fields in "time forward" to fields in "time past" with respect to an output sample point. No significant impairment is introduced by this time asymmetry.

Abstract: Cross talk between luminance and chrominance is eliminated in a Weston Clean PAL codec by replacing the band pass filter used in a conventional codec with a high pass filter (L) and introducing a compensating low pass filter (X) into the chrominance path, the filters being chosen such that L=4F'*(1-F') and X(f)=(F'(f):(1-F'(f)) where F'=F(2fsc-f) and L is a low pass (5.5 MHz) filter 248 applied before transmission in the PAL channel.

Abstract: In order to achieve increased resolution and to minimise cross-colour and cross-luminance effects in a PAL, NTSC or SECAM encoded signal, a high definition source (20) is pre-filtered at a higher line rate according to one of at least two methods (22, 24, 26). The filtered signal is sub-sampled (at 30, 32) at the source line standard where necessary, depending upon the degree of bandwidth reduction employed and one of the methods of pre-filtering and sub-sampling selected according (at 34) to a control signal (36) dependent on picture content. The luminance and chrominance components of the selected encoded signal are band pass filtered (46, 50, 52) and the chrominance signal modulated onto a sub-carrier (54) such that the frequency spectrum of the luminance and chrominance signals do not overlap.

Abstract: Missing lines of a video signal are interpolated from the signal on three successive fields by deriving low vertical frequency components solely from the central or current field and the higher vertical frequency components partly from the lines of the current field and partly from the lines of two adjacent fields. The total contribution from the current field is unity and the total contribution from each adjacent field is zero. The circuitry for deriving and combining these components comprises an arrangement of delays, adders, subtractors and multipliers (FIG. 7). The system can be used to generate the missing lines necessary to derive a sequentially scanned video signal from an interlaced scan video signal.

Abstract: A television test signal generator generates a test signal comprised of varying sinusoidal components each of which has a phase defined by the equation:.phi.=k.sub..phi. +k.sub.t .multidot.t+k.sub.t 2.multidot.(t.sup.2 /2) +k.sub.y .multidot.y+k.sub.yt .multidot.y.multidot.t+k.sub.y 2.multidot.(y.sup.2 /2) +k.sub.x .multidot.x+k.sub.xt .multidot.x.multidot.t+k.sub.xy .multidot.x.multidot.y+k.sub.x 2.multidot.(x.sup.2 /2)The use of more complex terms enables more sophisticated qualities of the television equipment to be gauged. The signal can be generated with the use of appropriate accumulators.

Abstract: Digital sampling of the luminance component of an N.T.S.C. color television signal is achieved by sampling with a mean frequency of twice the color subcarrier frequency to produce a sub-Nyquist sampled luminance signal. The sampling phase is changed between each pair of lines by an amount equal to half the interval between samples. The digital sampling can be applied to separate luminance and chrominance signals, to encoded N.T.S.C., and to the conversion of N.T.S.C. signals into signals of other form including PAL. Further, phase-perturbed sampling techniques permit the construction of an advantageous N.T.S.C. coder and de-coder system. The coder operates to cause spectrum folding of a luminance input signal by multiplying by a phase perturbed signal of twice the color subcarrier, and adding, the resultant being comb filtered to select frequencies which are multiples of the line frequency. The chrominance signals I, Q are combined into I.+-.

Abstract: In a small satellite earth station a directional antenna is rocked about a single axis, corresponding to an oscillatory change in declination, but constant hour angle. The rocking is approximately sinusoidal and has a period of one sidereal day. In a preferred arrangement the antenna is mounted on a rocking axis pivot which is fixed at right angles to, and rotatable for adjustment about, a polar axis member. The polar axis member is set up parallel to the earth's axis and the pivot is rotated about the polar axis member to set the hour angle. The rocking of the antenna is achieved by a crank and tie-rod arrangement driven by a clock motor mounted on the polar axis member. The arrangement is particularly simple and will, when set up, track a synchronous satellite without needing frequent adjustment.

Abstract: Reduction of noise in a television signal is achieved by comparing signals from successive frame or field scans and noting the differences. Low amplitude differences are assumed to represent noise and are attenuated. Differences above a predetermined level are assumed to represent movement and are not so attenuated by a factor K. Where there are inherent differences between the signals compared, for example they are different fields of an interlaced scan or are color signals with different subcarrier phases, the signal from the earlier scan is modified to compensate for these differences and provide a signal for comparison which is a better estimate of the expected signal on the later scan.

Abstract: PAL color television signals are processed in the form of a luminance signal Y and a composite chrominance signal consisting alternately of U + V on one line and U - V on the next. The composite chrominace signal is sampled at a frequency substantially equal to the PAL color subcarrier frequency and the luminance signal is sampled at the sub-Nyquist frequency of twice the color subcarrier frequency. Comb filters are included to remove undesired aliasing effects. The signals can be received in Y, U and V form or in encoded PAL form.