Abstract: Block matching is a robust and simple method of motion estimation for television pictures. One important parameter in block matching is the block size. Large blocks give more reliable motion estimation than small blocks, particularly in the presence of noise on the input picture, but they produce a coarser motion vector field. However, if the motion estimation is being used for motion compensated interpolation, for example the upconversion between 50 and 100 Hz display rates, the effects on picture quality of wrong vectors for whole blocks, and also of vectors that do not correctly follow the boundaries of moving objects, can be severe. According to the invention, large block (LB) matching is combined with the performance of more localized motion vectors to get pixel motion vectors. For any pixel, the motion vector will be one of four possibilities; the vector calculated for the block containing the pixel and the vectors (V1, V2, V3, V4) of the nearest blocks horizontally, vertically and diagonally.

Abstract: Block matching is well known as a robust and intuitively simple method of motion estimation for television pictures. However, when the purpose of the motion estimation is to provide accurate interpolation between images, conventional block matching suffers from the problem that the blocks are situated in the original fields rather than in the fields to be interpolated, leading to possible errors in the calculated motion field. The invention uses two-sided block matching. The search window is shared between the two input fields so that the candidate motion vectors related to be candidate pixel blocks all pass through the same points in the field to be interpolated. Thus, the "current block" becomes a notional area in the field to be interpolated, and each of the two search windows extends to half the maximum motion vector in each direction. Each motion vector points forward to the forward field by half its value and backward to the backward field by half its value.

Abstract: Motion vectors resulting from block matching are corrected for periodic structures by taking a more reliable vector from an edge of a moving object containing the periodic structure. This is done by calculating and comparing different error combinations to identify periodic structures and replacing a current motion vector with one of an adjacent pixel block either from the block to the left or from the block above, which ever yields the smaller error in the current block, or by taking a combination (e.g., mean) of both vectors. Advantageously, artifacts such as phase reversal of information in interpolated fields due to the presence of an erroneous vector in a periodic structure are reduced and very little additional processing is required after the block matching itself.

Abstract: An appearance of a double image around moving objects may be reduced or removed by generating motion compensated 100 Hz fields with the objects in correct positions. When, however, the results of this technique are applied to motion compensated upconversion, objects or parts of the picture which fall outside the range of measurable velocities appear on the upconverted display with a very strong double image, resulting from the display of information in the wrong temporal position generated by uncompensated or `fallback processing`. The subjective strength of the double image may be reduced by low pass filtering in a horizontal direction the components in the new fields of the upconverted signal which cause this double image.

Abstract: When film is transmitted by television, the same frame of film is used to generate both interlace television fields. This results in the information carried by the second field being temporally displaced from the original by 20 ms. This displacement gives rise to judder and double image artifacts in the received image. To overcome these defects the intermediate fields are generated from the original film frames by double sided motion compensated interpolation wherein the motion information is calculated using double sided block matching. The invention may be applied either in a television receiver or video signal decoder or in the television studio before the television signal is transmitted.

Abstract: Motion compensated video output signal (SO) produced by a soft switching or "fading" (SSW) between the a motion compensated interpolation (MCI) and a "fallback" mode (FFI) (e.g., intra-field), controlled by a "measure of confidence" (CF) in the reliability of the estimated motion vector. The confidence measure includes the following two components, namely: (i) a "basic confidence measure" (e.g., a block based, minimum motion estimation error (BME) and (ii) a measure based on the motion velocity or "speed" of movement of the motion vector or its horizontal (MVX) and vertical components (MVY).

Abstract: A method for signal equalization/ghost cancellation with an equalizing filter includes synthesizing a first equalizing filter function (LMS 1) with N (an integer) taps and associated coefficients. The respective coefficients are adjusted to minimize the error between a received reference signal output by the synthesized equalizing filter function and a non-distorted reference signal. The adjusted coefficients are evaluated to determine P (an integer) groups of M (an integer) taps and associated coefficients which contribute to said synthesized equalizing filter function according to a predetermined criterion and wherein P times M is less than N. A second equalizing filter function is synthesized corresponding to the first equalizing filter function with coefficients other than the P groups of M coefficients set to zero, and the P groups to M coefficients are adjusted to minimize the error.

Abstract: A method of measuring the noise in an active video image. During a given processing period, for each pixel of a video image considered to be stationary, the difference f.sub.d between the video signal corresponding to this pixel and the video signal corresponding t the pixel at the same spatial position in a preceding image is calculated. The square of the absolute value of the difference is calculated so as to obtain a value .vertline.f.sub.d .vertline..sup.2, and the value .vertline.f.sub.d .vertline..sup.2 is added to previously obtained values so as to form a sum value .SIGMA..vertline.f.sub.d .vertline..sup.2. The above operations are repeated until the number of pixels processed is equal to a given threshold value N.sub.0, or until the end of the processing period. At the end of the processing period, if the number of pixels processed is equal to the threshold value, the value .SIGMA..vertline.f.sub.d .vertline..sup.2 is used to determine a reference point giving a new noise-reduction coefficient k.sub.

Abstract: A system to facilitate mixing and recording video signals of different standards, e.g., in digital format via a digital video recorder, is disclosed. A first video signal in progressive scan format is converted to a standard format (e.g., line interlaced) prior to recording. A standard second video signal (e.g., line interlaced) is predistorted without format conversion prior to recording with the first signal to reduce artifacts associated with post-processing after recording. The predistortion involves vertically displacing selected picture elements. Before being displayed, a video output signal from the recorder is converted to the scan format of the first video signal.