Television – Image signal processing circuitry specific to television – Noise or undesired signal reduction
Reexamination Certificate
2000-03-06
2002-03-19
Lee, Michael (Department: 2711)
Television
Image signal processing circuitry specific to television
Noise or undesired signal reduction
C348S180000, C348S618000, C348S622000, C324S613000
Reexamination Certificate
active
06359658
ABSTRACT:
TECHNICAL FIELD OF THE INVENTION
The present invention is directed to an apparatus and method for enhancing a video signal and, more specifically, to an apparatus and method for measuring and filtering noise signals in an active video signal to create an enhanced video signal that produces a video image that is subjectively perceived to be superior to prior art video images.
BACKGROUND OF THE INVENTION
Video signal image enhancement circuitry in current television sets provides image enhancement by using noise measurement algorithms to measure and filter out noise signals. In the world of analog signals, the most common type of noise is Gaussian noise. Therefore, most prior art noise measurement algorithms are designed to measure and filter out only Gaussian noise.
Because noise on the video signal may arise from more than one source, noise that is detected in the video signal may be composed of more than one component. Each noise signal component may have its own characteristics. This is true even for noise signal components that are of the same type. For example, even if all the noise signal components in a video signal are of the Gaussian noise type, the frequency characteristics of each Gaussian noise signal component will differ from the frequency characteristics of each of the other Gaussian noise signal components.
With the level of digital transmissions in the environment continually increasing, digital transmission MPEG artifacts are more commonly appearing in video signals. To maintain high quality video signals, it is desirable to eliminate the effect of the MPEG artifacts upon the video signals. Because the existing noise measurement algorithms in current television sets can only measure Gaussian noise, it is necessary to add a separate MPEG artifact detector to the television set circuitry to detect, measure and eliminate MPEG artifacts from the video signal.
It would be desirable to have an apparatus and method that is capable of detecting more than one Gaussian noise signal component in a video signal. In addition, it would be desirable if the apparatus and method is also capable of detecting MPEG artifact noise signal components in a video signal.
To reduce noise in a video signal the noise level in the video signal must be identified and then subtracted from the video signal. A number of prior art techniques exist for identifying the noise level in a video signal. For example, a simple measure of the root-mean-square (“rms”) noise in a video signal may be obtained from the following equation:
N
estimated
=
∑
i
{
(
s
[
i
]
+
n
[
i
]
)
-
s
[
i
]
}
2
(
1
)
In Equation (1), N
estimated
is the estimated noise, s[i] is the signal without noise in the i
th
interval, and n[i] is the noise is the i
th
interval. Because the received signal is (s[i]+n[i]), the measurement of N
estimated
can only be obtained when s[i] is known. This suggests the possibility of measuring the noise in the horizontal blanking intervals (or the vertical blanking intervals) of the video signal where s[i] is known to be equal to a blanking level V
bl
. Although the blanking level V
bl
is not known exactly, it can be estimated as the long term average of (s[i]+n[i]) in the blanking interval.
Unfortunately, the estimate of the noise level in the blanking interval is not a reliable estimate of the noise level in the active video signal. This is because blanking signals are frequently reinserted by videocassette recorders and signal repeater stations in order to minimize clamp noise and sync jitter. That is, new blanking signals with less noise are inserted in place of the old blanking signals that may have more noise. The noise in the newly inserted blanking signals may therefore be less than the noise actually contained in the active video signal. Equating the noise level in the newly inserted blanking signals with the noise level in the active video signal would result in an underestimation of the actual noise level in the active video signal.
For this reason it is necessary to measure the noise level in the active video signal portion of the video signal. This introduces the problem of distinguishing between the signal and the noise in the active video signal. One approach for addressing this problem has been to assume that the image contains a certain minimum amount of horizontal stretches of constant luminance. In each of these stretches of L pixels having constant luminance (or almost constant luminance), it is assumed that variations within these stretches of pixels are caused by noise. It is possible to estimate the level of these local noise signals by determining the variance as follows:
N
estimated
=
(
x
,
y
,
f
)
=
∑
j
=
x
x
+
L
-
1
(
F
[
j
,
y
,
f
]
-
F
x
,
y
,
f
)
2
(
2
)
In Equation (2), a pixel position is specified by the coordinates (x,y,f). For a particular pixel, the value “x” specifies the position of the pixel in a line, the value “y” specifies the position of the line in a frame, and the value “f” specifies the position of the frame. During a broadcast the pixels are transmitted sequentially. Therefore, the location of any particular pixel during a transmission may be specified by a single (i.e., one dimensional) coordinate. The single coordinate is referred to as “i” and the value of “i” is calculated by:
i=x·T
x
+y·T
L
+f·T
f
(3)
In Equation (3), T
x
is the sample time, T
L
is the line duration and T
f
is the field duration. The values for T
x
, T
L
and T
f
are fixed for a particular standard (e.g., PAL, NTSC). The location of a pixel in a transmission may be specified using this method.
In Equation (2), N
estimated
(x,y,f) is the estimated noise, L is the number of pixels in a selected stretch of pixels, and F[x,y,f] is equal to (Y[i]+n[i]). Y[i] is the luminance in the i
th
interval and n[i] is the noise is the i
th
interval. In Equation(2), F
x,y,f
is the local average of the (Y[i]+n[i]) signal and is calculated by:
F
x
,
y
,
f
=
1
L
∑
k
=
x
x
+
L
-
1
F
[
k
,
y
,
f
]
(
4
)
To utilize the variance method to estimate the noise level in an active video signal it is necessary to calculate the variance from a large number of areas. It is assumed that the image contains a certain amount of small areas of constant luminance. It is also assumed that the areas yielding the smallest variance contain no detail from the image but only noise. The problem is that if the image contains a lot of “flat” area (where there is no contrast or very little contrast in the image), the variance method leads to an underestimation of the noise. This makes the noise measurement dependent upon the picture content.
One method for solving this problem is to take the average of the noise estimates over the R smallest noise estimates where R is a preselected number that is a non-zero positive integer. The averaging of the noise estimates decreases the dependency of the noise measurement upon the picture content.
After a noise measurement system estimates the amount of noise in a video signal, the noise measurement system sends the noise estimate to other signal processing elements of the video system. One such signal processing element is a noise subtraction element that is capable of subtracting the noise components of the signal from the active video signal. The subtraction of the noise components from the active video signal provides an enhanced active video signal that is capable of producing improved video images with less noise content.
The presence in a video signal of noise components having differing frequency characteristics may cause a noise measurement system (and its associated noise subtraction element) to make corrections to the video signal that do not provide the highest quality image possible from the viewpoint of subjective perception. The subjective perception of viewers of the video image
He Haiyan
Hentschel Christian
Gathman Laurie E.
Lee Michael
Philips Electronics North America Corporation
Tran Trang U.
LandOfFree
Subjective noise measurement on active video signal does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Subjective noise measurement on active video signal, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Subjective noise measurement on active video signal will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2865624