Image analysis – Image enhancement or restoration – Artifact removal or suppression
Reexamination Certificate
2000-06-23
2004-06-29
Mehta, Bhavesh M. (Department: 2625)
Image analysis
Image enhancement or restoration
Artifact removal or suppression
C382S274000, C382S282000, C382S156000, C382S159000, C358S003260, C358S453000, C358S463000
Reexamination Certificate
active
06757443
ABSTRACT:
TECHNICAL FIELD
The present invention relates to a picture processing apparatus, a picture processing method, a learning apparatus, and a learning method that have a function for removing noise from an input picture signal.
RELATED ART
In a picture signal processing system or the like used in for example a television receiver, to remove noise from a picture signal, a motion adaptive recursive filter has been used.
FIG. 12
shows an example of the structure of the motion adaptive recursive filter. Referring to
FIG. 12
, an input picture signal of the current frame is supplied to a subtracting circuit
100
. In addition, a picture signal of the chronologically earlier frame stored in a frame memory
99
is supplied to the subtracting circuit
100
. The subtracting circuit
100
generates the difference between the picture signal of the current frame and the picture signal of the chronologically earlier frame and supplies the generated difference to an absolute value calculating circuit
101
. The absolute value calculating circuit
101
converts the supplied difference into an absolute value. Output data of the absolute value calculating circuit
101
is supplied to a threshold value processing circuit
102
. The threshold value processing circuit
102
performs a threshold value process for the supplied absolute value with a predetermined threshold value and determines whether or not each picture has a motion.
The determined result of the threshold value processing circuit
102
is supplied to a weighting value generating circuit
103
. The weighting value generating circuit
103
designates a weighting value k corresponding to the determined result of the threshold value processing circuit
102
. The weighting value k is supplied to amplifiers
104
and
105
. The amplifier
104
adjusts the amplitude of the input frame. The amplifier
105
adjusts the amplitude of the chronologically earlier frame stored in the frame memory
99
. The amplifier
104
amplitudes the input signal by k times. The amplifier
105
amplitudes the input signal by (1−k) times.
When the threshold value processing circuit
102
has determined that the current pixel does not have a motion, the value k is designated to a predetermined fixed value in the range from 0 to 0.5. An adding device
106
is disposed downstream of the amplifiers
104
and
105
. The adding device
106
outputs the added value of an output value of the amplifier
104
(the pixel value of the current pixel of the current frame is multiplied by k) and an output value of the amplifier
105
(the pixel value of the current pixel of the chronologically earlier frame is multiplied by (1−k)). The current pixel position of the current frame is the same as the current pixel position of the chronologically earlier frame. On the other hand, when the threshold value processing circuit
102
has determined that the current pixel has a motion, the value k is designated to “1”. Thus, the adding device
106
directly outputs the pixel value of the current pixel of the current frame.
However, in the related art reference, there are problems (a) to (d) that follow. As the problem (a), since a non-moving portion is weighted by a predetermined weighting value (fixed value k), when there are chronological fluctuations of the noise level and the relation between the noise level and the signal level, the noise cannot be accurately removed. As the problem (b), when the noise level is large, since a non-moving portion is incorrectly determined as a moving portion, the noise removing effect deteriorates. As the problem (c), when a moving portion is incorrectly determined as a non-moving portion, a dimmed portion may take place. As the problem (d), noise cannot be removed from a moving portion.
Therefore, an object of the present invention is to provide a picture processing apparatus, a picture processing method, a learning apparatus, and a learning method that allow noise to be accurately removed in an environment that the noise level chronologically fluctuates.
DISCLOSURE OF THE INVENTION
In one aspect of the present invention, a picture processing apparatus includes a storing portion for storing a plurality of frames of an input picture signal, a pixel extracting portion for extracting at least one pixel from a considered frame stored in the storing portion and at least one pixel from the other frames stored in the storing portion, a noise detecting portion for detecting chronological fluctuation of noise level between the frames according to the pixels extracted from the frames by the pixel extracting portion, and a picture signal generating portion for processing the considered frame of the input picture signal according to the fluctuation of the noise level and thereby generating a picture signal from which noise is removed.
A picture processing method of the present invention includes the steps of extracting at least one pixel from a considered frame of a plurality of frames of an input picture signal and at least one pixel from the other frames, detecting chronological fluctuation of noise level between the frames according to the pixels extracted from the frames, and processing the considered frame of the input picture signal according to the fluctuation of the noise level and thereby generating a picture signal from which noise is removed. frames according to the pixels extracted from the frames, and processing the considered frame of the input picture signal according to the fluctuation of the noise level and thereby generating a picture signal from which noise is removed.
In another aspect of the present invention, a learning apparatus includes a noise adding portion for adding a noise component to a teacher picture signal so as to generate a student picture signal, a storing portion for storing a plurality of frames of the student picture signal that is output from the noise adding portion, a pixel extracting portion for extracting at least one pixel from a considered frame of the student picture signal stored in the storing portion and at least one pixel from the other frames of the student picture signal stored in the storing portion, a noise detecting portion for detecting chronological fluctuation of noise level between the frames according to the pixel of each frame extracted by the pixel extracting portion and generating class information according to the fluctuation of the noise level, and a predictive coefficient calculating portion for calculating predictive coefficients for generating an output picture signal having the same quality as the teacher picture signal with an input picture signal having the same quality as the student picture signal according to the class information, the teacher picture signal, and the student picture signal.
A learning method of the invention comprises the steps of adding a noise component to a teacher picture signal so as to generate a student picture signal, extracting at least one pixel from a considered frame of the student picture signal that has been stored and at least one pixel from the other frames of the student picture signal that has been stored, detecting chronological fluctuation of the noise level between the frames according to the pixel of each frame extracted at the extracting step and generating class information according to the fluctuation of the noise level, and calculating predictive coefficients for generating an output picture signal having the same quality as the teacher picture signal with an input picture signal having the same quality as the student picture signal according to the class information, the teacher picture signal, and the student picture signal.
According to the present invention, corresponding to pixel data extracted from a predetermined number of frames, a noise removing process that accurately corresponds to the fluctuation of a noise component between frames can be performed.
REFERENCES:
patent: 4837618 (1989-06-01), Hatori et al.
patent: 4926361 (1990-05-01), Ohtsubo et al.
patent: 4999704 (1991-03-01), Ando
patent: 5289274 (1994-02-01), Kondo
Hamamatsu Toshihiko
Kondo Tetsujiro
Nakaya Hideo
Okumura Yuji
Shiraki Hisakazu
Frommer William S.
Frommer & Lawrence & Haug LLP
Kassa Yosef
Mehta Bhavesh M.
Polito Bruno
LandOfFree
Picture processing apparatus, picture processing method,... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Picture processing apparatus, picture processing method,..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Picture processing apparatus, picture processing method,... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3353534