Television – Image signal processing circuitry specific to television – Noise or undesired signal reduction
Reexamination Certificate
1997-09-19
2001-03-13
Peng, John K. (Department: 2714)
Television
Image signal processing circuitry specific to television
Noise or undesired signal reduction
C348S624000, C348S909000, C348S627000, C358S533000
Reexamination Certificate
active
06201582
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The subject invention relates to video display apparatus and, in particular, to suppressing moiré disturbances in the displaying of video signals on the video display apparatus.
2. Description of the Related Art
Moiré is a term commonly used to describe disturbances on a video display which look like waves on water. A common example in the real world is the interference between two fences. Moiré appears when two sampling processes do not match each other, and there is insufficient filtering.
The description of moiré on video monitors is complex, because of the two dimensional nature and the completely different sampling structure of the screen of the monitor and that of the video signal. An easy way to describe the origins is given in the one-dimensional time and frequency domain.
FIGS. 1A-1C
show the structure of a sampling process of an incoming continuous signal. As shown in
FIG. 1A
, the sampling process may be done by an ideal analog-to-digital converter. The output data s(n) (
FIG. 1C
) are time discrete samples of the input signal s(t) (
FIG. 1B
) with a spacing of 1/f
s
, where f
s
is the sampling frequency. The spectrum of the input signal is repeated by the sampling frequency and all multiples of it. These frequencies will be called carriers, because their behavior is comparable to an amplitude modulation.
Moiré disturbances are caused by alias frequencies and beat frequencies. Alias frequencies appear when the repetition spectrum of the first carrier overlaps with the baseband. In this case, high signal frequencies cause low repetition frequencies. Disturbances caused by aliasing are not removable without loss of signal information, when the alias frequency occurs within the baseband bandwidth. The usable baseband is limited by the Nyquist frequency, which is half the sampling frequency.
Beat frequencies near the Nyquist limit are shown in
FIGS. 2A-2C
. The input signal (
FIG. 2A
) is below the Nyquist limit and the sampling process gives an additional frequency line at f
s
−f
o
=0.53f
s
. The sampled signal (
FIG. 2B
) has a period of 1/f
o
=1/(0.47fs), which is the signal frequency. Due to the fact that the signal frequency and the repetition frequency are close together, as shown in
FIG. 2C
, the frequency difference f
b
=¦f
o
−(f
s
−f
o
)¦=¦f
s
−2f
o
¦ can be seen as a beat frequency causing modulation in the sampled signal. The beat frequency is just the frequency difference of two physical frequencies, therefore, the beat frequency itself is not a physical frequency. To remove a beat frequency, at least the higher physical frequency must be suppressed. Unfortunately, beat frequencies often come along with alias frequencies.
FIGS. 3A-3C
show a time discrete (or digital) low-pass filter along with the input signal spectrum ¦Si(f)¦ and the output signal spectrum ¦So(f)¦. An important property of digital filters is the symmetrical frequency response to the Nyquist frequency. Low frequency aliasing cannot be suppressed without significant loss of signal information. In some cases, when the cutoff frequency is below the Nyquist limit, high frequency aliasing can be suppressed along with disturbing beat frequencies. Unfortunately, high signal frequencies would be suppressed in the same way. Beat frequencies can be removed by a digital low-pass filter when the cutoff frequency is significantly below the Nyquist limit.
The two-dimensional sampling process of the video signal creates carrier amplitudes at multiples of the video format, or so-called “resolution”. For instance,
FIG. 4
shows the frequency space of a graphics standard in the format 1600×1200 pixels (N
x
·N
y
) The first carrier frequency is determined by the distance of two pixels. Therefore, the first carrier amplitudes become 1/N
x
and 1/N
y
, which is 1600 cy/pw and 1200 cy/ph (1600 cycles per picture width, and 1200 cycles per picture height).
As shown in
FIG. 4
, the Nyquist limit has a rectangular shape with borders at half of the sampling frequency. The video format 1600×100 pixels has a maximum resolution of 800 cy/pw in the horizontal, and 600 cy/ph in the vertical directions. The three patterns describe the limits at the Nyquist limit. In the horizontal direction, the Nyquist limit is given by alternating pixel amplitudes, while in the vertical, the Nyquist limit is given by alternating line amplitudes, and in the diagonal direction, by alternating pixel and line amplitudes, resulting in a checkerboard pattern.
Video moiré appears mostly at the Nyquist limit. The two patterns, checkerboard and alternating pixels (see FIG.
4
), are most critical. Both have, in common, the horizontal Nyquist limit. Therefore, it is sufficient to suppress only the area around the horizontal Nyquist limit. In most cases, a two-dimensional low-pass filter is not needed, a one-dimensional horizontal low-pass filter is sufficient.
Linear low-pass filters have several disadvantages. The resolution will decrease and high contrast transitions will be muted resulting in the picture appearing less sharp. Also single lines or fine details will be seriously suppressed, and the picture impression becomes weak. Well-known non-linear filters, e.g., median filters, can preserve sharp edges, but small details will still be suppressed. Additionally, these filters create a certain amount of alias disturbances.
SUMMARY OF THE INVENTION
In general, it is an object of the invention to provide a non-linear filter which suppresses the Nyquist frequency, while leaving impulses and transitions unchanged. For the common case of video moiré at high horizontal input frequencies, it is desirable to limit the filter to a one-dimensional filter, operating in the horizontal direction.
The above object is achieved in a circuit for suppressing a Nyquist frequency in a video signal, said circuit comprising an input for receiving an input video signal; means coupled to said input for low-pass filtering said input video signal; means also coupled to said input for detecting a Nyquist frequency in said input video signal; and means, having inputs coupled to said input, said low-pass filtering means and said detecting means, for mixing said input video signal and an output of said low-pass filtering means in response to an output from said detecting means.
In a preferred embodiment of the invention, the detecting means comprises an input for receiving said input video signal; means coupled to said input for high-pass filtering said input video signal; means coupled to said input for bandpass filtering said input video signal; first and second means for forming an absolute value coupled, respectively, to outputs of said high-pass filtering means and said bandpass filtering means; and means coupled to outputs of said first and second absolute value forming means for forming a blending factor.
FIGS. 5B-5D
show examples for suppressed and unchanged patterns by use of a non-linear filter for Nyquist frequency suppression. The decision is done within a processing window of 5 pixels in length (see FIG.
5
A). The minimum length for this filter is 4 pixels, but there is no fixed maximum limit.
The advantages of the filter of the subject invention over linear filters, median filters, and other known non-linear filters are:
no loss of sharpness at transitions;
impulses remain with full amplitude (i.e., no degradation);
only alternating patterns, which result in moiré (beat and alias frequencies), will be suppressed;
usual video signals having a bandwidth limitation below the Nyquist frequency will almost be not affected; and
small hardware or software cost.
REFERENCES:
patent: 4065784 (1977-12-01), Rossi
patent: 4823186 (1989-04-01), Muramatsu
patent: 4876591 (1989-10-01), Muramatsu
patent: 5175621 (1992-12-01), Maesato
patent: 5177611 (1993-01-01), Gibson et al.
patent: 5231677 (1993-07-01), Mita et al.
patent: 5313301 (1994-05-01), Lee
patent: 5321513 (1994-06-01), Kondo et al.
patent: 5379074 (1995-01-01), Hwang
patent: 5400083 (1
Desir Jean W.
Goodman Edward W.
Peng John K.
Philips Electronics North America Corporation
LandOfFree
Circuit for moiré suppression does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Circuit for moiré suppression, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Circuit for moiré suppression will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2466565