Abstract: An MPEG2 encoded signal comprising intra-encoded pictures I and inter-encoded pictures B is decoded (70-74) to baseband and re-encoded (76) to I-pictures. The quantisation levels Q.sub.v and the motion vectors M.sub.v of the B pictures are retained unchanged and maintained in association with the corresponding pictures re-encoded as I-pictures. The I-pictures can be decoded to baseband, processed, and re-encoded as I-pictures without degradation. The I-pictures can then be re-encoded back to e.g. I, B pictures with minimal degradation. Alternatively, some of the I-pictures are decoded and converted to B pictures retaining unchanged the quantisation levels thereof.

Abstract: A frame of time-domain video signals Si is supplied via a frame reorderer (40) to a motion vector estimator (41), and then to a DCT transformer (43) and a quantizer (44) to produce a quantized, transformed current frame S.sup.T i/q. A subtracter (42) subtracts a transformed quantized prediction frame S.sup.T p/q from S.sup.T i/q to produce an inter-encoded signal .DELTA.. The prediction frame S.sup.T p/q is produced by adding the prediction signal to S.sup.T p/q to the inter-encoded signal .DELTA., dequantizing (46) and inverse transforming (47) the sum and applying the sum to a motion compensation circuit (49) to produce a time domain prediction signal Sp which is transformed (50) and quantized (51) to produce S.sup.T p/q The quantizers (44) and (51) are controlled by the same quantization level Qv. The encoder and a corresponding decoder allow frames to be encoded and decoded with reduced reconstruction errors in systems where compress ion codecs are cascaded.

Abstract: A video data compression system is described utilizing frequency separation and quantization. The video data is separated into primary frequency components (components 4, 5, 6) and secondary frequency components (components 0, 1, 2, 3). The primary frequency components undergo fewer frequency separating stages (N>M) than the secondary frequency components. The quantization step width (Q.sub.1, Q.sub.2) applied to the differing primary and secondary components is varied in dependence upon the relative human visual responsiveness (R(f.sub.s)) to the spatial frequency (f.sub.s) represented by that component. In addition, those frequency components that have undergone fewer frequency separating stages are subject to a larger quantization step width than the other frequency components since the fewer frequency combination stages such frequency components will subsequently undergo has the result that the image degradation introduced by quantization is less noticeable.

Abstract: A method or apparatus for processing an input frame of image data comprising converting the input frame into a plurality of output frames, the spatial frequency spectrum of each of said plurality of output frames being that of a respective predetermined frequency band of the spatial frequency spectrum of the input frame and applying separately to at least one of said plurality of output frames a respective gain.

Abstract: Decoder reconstruction error is estimated so as to provide compensation at the encoder which equalizes the errors over different frames, since the errors created by some frames are lower than those created by other frames. The compensation is provided by means of temporal weighting applied in the form of temporal bands each associated with specific numbered frames within a frame group. The technique can be applied to a logarithmic temporal decimation frame structure in which one frame within a group of 2.sup.n frames is an `I` frame and the remaining frames are `B` frames.

Abstract: An apparatus for recording video data with a multi-level arrangement of error correction coding. Input video data is demultiplexed so as to form a plurality of data streams each having a plurality of data blocks. A stream error correcting code is generated for each of the plurality of data streams, which is adaptable to allow correction of a data block. The respective data blocks and stream error correcting code of each data stream is distributed or arranged in a predetermined manner. The distributed data blocks and stream error correcting code of each data stream are recorded on recording tracks of a recording medium by n recording heads such that only either a data block or a stream error correcting code of a respective data stream is recorded at corresponding positions along n adjacent recording tracks of the recording medium.

Abstract: A method and apparatus for processing an input image by converting an input frame of a predetermined number of pixels into a plurality of output frames, the spatial frequency spectrum of each of said plurality of output frames being that of a respective predetermined frequency band of the spatial frequency spectrum of the input frame, the processing including filtering the input frame in one direction so as to produce a first output frame of said predetermined number of pixels, the first output frame having only one of the high spatial frequencies and the low spatial frequencies of said input frame in said one direction, and subtracting said first output frame from said input frame so as to produce a second output frame of said predetermined number of pixels, the second output frame having the other of the high spatial frequencies and the low spatial frequencies of said input signal in said one direction.

Abstract: A motion compensated video processing method includes encoding a video frame group I.sub.0 -I.sub.7 into one in which one frame I.sub.0 of the group is a basic input frame or `I `frame, and the other frames B.sub.1 -B.sub.7 are derived from calculating the difference between a current frame and two neighbouring frames so as to form `B ` frames. The `B ` frames are derived in a logarithmic progression. The `B ` frames are quantized by applying selective quantization weighting to predetermined input frames. Encoding can be of the forms described as open loop or closed loop, preferably open loop. Decoding can be achieved by applying inverse quantization weighting to the input frames prior to selective summnation in order to achieve a constant reconstruction error for each frame even though quantization errors carry over from each reconstruction stage.

Abstract: A data compression system decorrelates input data into sets of frequency component data. Different sets of frequency component data are entropy encoded using different coding tables matched to the data values characteristic of the sets of frequency component data they are used to compress. Matching the coding tables to the sets of frequency component data improves compression. The order in which the data values from within each set of frequency component data is fed to the entropy encoder is varied to exploit any intra-set correlation.

Abstract: A motion compensated video processing method includes encoding a video frame group I.sub.0 -I.sub.7 into one in which one frame I.sub.0 of the group is a basic input frame or `I` frame, and the other frames B.sub.1 -B.sub.7 are derived from calculating the difference between a current frame and two neighboring frames so as to form `B` frames. The `B` frames are derived in a logarithmic progression. The `B` frames are quantized by applying selective quantization weighting to predetermined input frames. Encoding can be of the forms described as open loop or closed loop, preferably open loop. Decoding can be achieved by applying inverse quantization weighting to the input frames prior to selective summation in order to achieve a constant reconstruction error for each frame even though quantization errors carry over from each reconstruction stage.

Abstract: A video frame group is encoded into `I` and `B` frames, and the `I` frames are re-ordered, such as by substituting each `I` frame into the next frame group, so that the timing delay upon decoding can be reduced compared to the decoding delay for a non-re-ordered frame group. The `B` frames may also be re-ordered within each group.

Abstract: In video motion vector generation and compensation, it is common for the chrominance sampling rate to be half the luminance sampling rate. When only a luminance motion vector is used, there can be rounding problems if either the horizontal or vertical motion vector (or both) is odd valued and the luminance motion vector is divided by two to generate a chrominance motion vector. For this reason, the luminance motion vector is provided with an additional bit for either or both axes if odd valued, and a rounding decision, if necessary, is made on the basis of the digital value of the additional bit.

Abstract: A video signal coding apparatus for coding first and second input video images comprises a motion vector generator for matching pixels of the second input image B with pixels of the first input image A to derive motion vectors representing the motion of pixels in the second image between the two images, each motion vector being associated with the location of a pixel in the second image B and the location of the matched pixel in the first image A. The apparatus includes an encoder for processing the two input images A, B to produce first and second output video images C, D, pixels of the first output image C being produced by addition of equally weighted pixels in the two input images A, B, and pixels of the second output image D being produced by subtraction of equally weighted pixels in the two input images A, B.

Abstract: Dropout is compensated in a video signal that has samples comprising respective different color information components in alternate lines and samples comprising luminance information time multiplexed with the color samples in each line. Adders (78, 82) average first and second sample (S1, S2) positioned one line after and one line before a central reference sample SR to produce a first averaged sample SA1, and third and fourth samples (S3, S4) positioned two lines after and two lines before the sample SR to produce a second averaged sample SA2. Gating means (48 to 60) outputs any one of the sample SR, the averaged samples and the first to fourth samples.