This application is a National Phase Application of PCT International Application No. PCT/GB03/02348, International Filing Date May 29, 2003, entitled “VIDEO SIGNAL PROCESSING”, which in turn claims priority from Great Britain Patent Application, 0212430.3, filed May 29, 2002, which are both incorporated by reference herein in their entirety.
This invention concerns the sampling or re-sampling of television signals to achieve an orthogonal sample structure.
Since the adoption of ITU/CCIR Recommendation 601 the digital processing of television signals has almost always been carried out with an orthogonal sampling structure. However, there are applications in which it is difficult to convert the signal into this form; a well-known example is the processing of the output of a domestic video cassette recorder. In this case the instability of the line frequency makes it difficult to derive an orthogonal sampling grid.
In order to sample an analogue television signal orthogonally, it is necessary to derive a repetitive sampling clock signal which defines the instants at which the signal is to be digitised. If the signal were stable these instants would be regular, and the clock would be an integral multiple of the signal's line frequency. In the case of an unstable signal the timing of the clock pulses must vary, so that the same number of samples is taken on every line, even though all lines are not of the same duration.
One known method of doing this is to use a high-bandwidth (of the order of 1 khz), frequency-agile phase locked loop to derive the clock. The clock oscillator attempts to follow the signal's varying line frequency and so maintain the same number of samples on every line. This is difficult to achieve; one problem is that the wide-bandwidth loop is more susceptible to noise than a narrow-band loop would be. Another problem is that if the loop is capable of operating over a wide frequency range it is less able to phase-lock accurately because of practical limitations to the resolution of the control signal.
A second known method is to sample the signal with a stable, free-running sampling clock and then process the resulting non-orthogonal samples to obtain orthogonal samples. This is done by the horizontal interpolation of new sample values from the actual samples taken. This interpolation corresponds to a filtering process which modifies the frequency response of the sampled signal; if a flat response is required a large number of input samples need to be used to generate each interpolated output sample.
The invention seeks to overcome these difficulties of the known methods and to provide a technique for obtaining orthogonally-sampled digital video from both stable and unstable source material.
The invention consists, in one aspect, of a method of generating an orthogonally sampled video signal, in which the signal is sampled by a substantially constant frequency sampling clock locked to the average line frequency of the video signal, and the resulting samples are shifted to an orthogonal sampling grid.
Preferably the samples are shifted without interpolation between samples when the instantaneous input line frequency is equal to the average input line frequency.
Suitably no more than three input samples are used to generate an output sample.
Advantageously the required sample shift is derived from a line phase measure, characterised in that the measurement is made in integer sample interval units and interpolated to a precision corresponding to a fraction of a sample interval.
In an alternative embodiment the samples are delayed prior to shifting so that a line phase measure corresponding to the phase of a later sample, is used to calculate the required shift for an earlier sample.
In a further aspect of the invention there is provided a method of generating an orthogonally sampled video signal in which the signal is sampled by a sampling clock and the resulting samples are shifted to an orthogonal sampling grid, wherein the required sample shift is derived from a line phase measure.
Two examples of the invention will now be described with reference to the drawings in which:
Referring to
The output samples from the ADC (2) are passed to a digital sync-separator (4) which calculates, to sub-sample accuracy, the position of the horizontal timing reference point (normally the mid-point of the leading edge of the line sync pulse) of each line of the input video signal (1) with respect to the sampling clock from the oscillator (3). A sync phase comparator (5) derives the phase of the detected timing reference point relative to a line frequency signal obtained by dividing (6) the sampling clock frequency by the required number of orthogonal samples per line. The phase difference output of this comparator would typically be expressed as the sum of an integer number of clock samples and a fraction of a sampling period.
The output of the phase comparator (5) is used for two different purposes. Firstly, the phase difference is low pass filtered by a loop filter (7) and used to control the frequency of the oscillator (3). The closed-loop bandwidth of the control loop is low, typically in the range 0.1 to 50 Hz, and so, for stable input signals, the sampling structure will be orthogonal and accurately phased to the horizontal time reference point. In the case of an unstable input the samples will not be orthogonal, but the average number of samples per line will be a multiple of the average input line frequency.
Secondly, the output from the phase comparator (5) controls a sample shifter (8) which converts the non-orthogonal samples from the ADC (2) into an orthogonally sampled output (10). (The phase comparator output may be filtered or temporally up-sampled as will be explained later.)
The sample shifter (8) must be able to shift samples by the maximum phase error between the stable line frequency signal from the divider (6) and the input line frequency. Methods of shifting by an integer number of samples are well known, typically a buffer memory may be used. Provision is also made for shifting the samples by a fraction of a sampling interval. As has been mentioned in the introduction to this specification, it is complicated to do this while maintaining a good frequency response. In the invention a very simple interpolator can be used; typically only two consecutive input samples are used to derive an output sample lying between them. This simplification is possible because the phase control loop for the oscillator (3) is designed so that, for stable inputs, the fractional part of the output of the comparator (5) is zero and so no Interpolation (and no consequent distortion) occurs to such inputs. The inventors have appreciated that unstable video signals (such as those from VHS playback) most commonly have existing distortion (such a reduced frequency response) and thus the additional distortion due to low-order interpolation will be insignificant, and not apparent to the viewer.
In an embodiment, there is some thresholding employed in the shifting circuit, such as a coring function, in order to ensure that for very low phase differences, no interpolation at all is performed. In such cases, the error in the output introduced by interpolating is greater than that introduced by not interpolating (and thus not shifting samples to exactly the “correct” location).
As mentioned above the output from the comparator (5) can be processed to obtain further benefits; these further embodiments of the invention will now be described with reference to
The phase comparator (5) can only produce an output once per input line sync pulse; however, the phase relationship between the sampling clock from the oscillator (3) and an ideal orthogonal sampling structure may be continually changing and, ideally, the control signal to the sample shifter (8) should follow the changes so that each sample is correctly shifted to an-orthogonal position. Acceptable results can be obtained by using the same shift for all the samples of one output line, but visible picture width changes may occur as the input line frequency changes.
The interpolating filter may also be used to increase the resolution of the control signal; for example, only the integer part of the phase error could be input to the filter (20) whereas the filter output could have a much higher resolution, typically five fractional-significance bits. If there is some dither on the phase measurement, perhaps due to sync noise or even the changing value of the actual phase error, then increased phase resolution will be obtained. This use of several phase error measurements to give a more accurate estimate of the error may be used alone, or in combination with the up-sampling feature of the filter.
As well as lacking temporal resolution, the output from the comparator (5) is inherently “backward looking”; it has no knowledge of the future. This deficiency can be overcome be inserting a compensating delay (21) in series with the video input to the sample shifter (8). This delay can be chosen in conjunction with the interpolation characteristic of the filter (21) so that the shift value applied to a particular sample is derived taking into account the phase of samples within a filter aperture extending before and after the current sample.
The design of the filter (20) should be optimised to achieve the best prediction of the required sample shift and rejection of sync noise, taking into account expected video characteristics due to such factors as tape transport dynamics and head-switching transients. The filter may, for example, have non-linear characteristics.
Various other implementations of the basic concept will be apparent to those skilled in the art, for example separate phase comparators could be used for deriving the oscillator control signal and the interpolator control signal, each optimised for its particular application. The oscillator control loop may have a non-linear amplitude characteristic to minimise the phase error for stable signals. Though embodiments of the invention have been described with analogue input, others may be implemented with digital input where the sampling operation of the ADC is replaced by a digital re-sampling operation.
Number | Date | Country | Kind |
---|---|---|---|
0212430.3 | May 2002 | GB | national |
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/GB03/02348 | 5/29/2003 | WO | 00 | 11/16/2005 |
Publishing Document | Publishing Date | Country | Kind |
---|---|---|---|
WO03/103274 | 12/11/2003 | WO | A |
Number | Name | Date | Kind |
---|---|---|---|
4731646 | Kliem | Mar 1988 | A |
5062005 | Kitaura et al. | Oct 1991 | A |
5150201 | Mehrgardt et al. | Sep 1992 | A |
5280352 | Herrmann | Jan 1994 | A |
5841480 | Rhodes | Nov 1998 | A |
6100661 | Stessen et al. | Aug 2000 | A |
6297849 | Stessen et al. | Oct 2001 | B1 |
6323913 | Prange | Nov 2001 | B1 |
Number | Date | Country |
---|---|---|
0 477 823 | Sep 1991 | EP |
0 865 198 | Sep 1998 | EP |
WO 03103274 | Dec 2003 | WO |
Number | Date | Country | |
---|---|---|---|
20060087589 A1 | Apr 2006 | US |