Image analysis – Image enhancement or restoration – Image filter
Reexamination Certificate
1998-05-07
2001-01-16
Lee, Thomas D. (Department: 2724)
Image analysis
Image enhancement or restoration
Image filter
C382S254000
Reexamination Certificate
active
06175657
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to an adaptive filtering system to reduce the level of Gaussian noise in a memory field being scanned by rows. More particularly, the system of the invention is adapted to reduce the noise of an image to be displayed on a display screen by implementing an adaptivity in response to brightness changes, according to a predefined sensorial response system.
BACKGROUND OF THE INVENTION
Noise is inherent to any communication system and, in television broadcasting networks, originates in different ways. At the source of image, noise may be either generated by photon or thermal noise in a video camera or by granular noise in films of cameras. Signal processing circuits and recorder play back systems add noise because the signal circulates in an intrinsically noisy loop.
Finally, the signal may be carried to the transmitter via a distribution network that adds more noise, and broadcast via an intrinsically noisy transmission medium. The receiver circuits themselves add noise. Many techniques have been developed to reduce the problem of Gaussian noise and these are described in literature. The adaptivity of noise level to a certain image/video sequence is a key feature of the so-called “smart” filters.
Generally, this adaptivity is obtained using non-linear techniques, such as averaging filters or filters with variable behavior depending on the estimation of some parameters. For instance, many known filters are obtained using scale and position estimates, or by estimating the temporal correlation (Recursive Averaging Filters) which is the most common techniques presently used in TV environments.
These known techniques have intrinsic disadvantages and/or limitations. In the case of a video sequence strongly affected by noise, the filter may confuse the noise with motion effects resulting in a poor noise reduction effect. To improve this situation, the user could automatically select from among a set of preordered response characteristics, such as to dynamically accentuate or reduce the filtering effect. However, these systems of arbitrary selection influence the temporal frequency, introducing a sort of “comet effect” which may become more annoying than an excessive edge smoothing. On the other hand, motion detection requires at least one field memory in the recursive loop of the filter with the consequent increment of total memory requirement for the system.
SUMMARY OF THE INVENTION
It is an object of the present invention to provide a Gaussian noise filtering system that minimizes the memory requirement for noise filtering and that introduces an adaptivity to the local picture characteristics based on the sensorial response to brightness changes of the so-called Human Visual System without penalizing edge definition by an undue smoothing. The filtering system of the invention provides for an effective response to these important requirements by processing data relative exclusively to a certain memory field, thus minimizing the memory requirement. The essential characteristic of the system of the invention is the utilization of a fuzzy logic process which allows for a current definition of the adaptivity parameters bearing in mind the characteristics of a certain scheme of sensorial response, such as, for example, the scheme of response to brightness changes known as the Human Visual System (HVS). The method of the invention reduces the noise affecting a certain area of a field scanned by rows, without employing information determined from a preceding field or picture which, in contrast, is the classic approach of the known systems wherein the noisy pixel is substituted with an averaged pixel value of the nearby ones belonging to a certain processing window. This known approach has the serious disadvantage of smoothing the edges and the contours of the objects thereby depressing image definition.
In contrast, according to the present invention, each noisy pixel singled out within a certain processing window, is substituted with a weighted average of the surrounding pixels. This is done according to a preestablished degree of similarity between the same pixel to be substituted and each of the nearby ones.
In this manner, if K
i
provides a measure of the degree of similarity between the pixel X to be filtered and a generic pixel X
i
belonging to the process window, the output filtered pixel value is given by:
X
(
i
)
out
=K
i
*X+
(1
−K
i
)*
X
i
(1)
If K
i
is equal to 0, X and X
i
are equal and thereby X may be substituted with X
i
. Contrarily, if K
i
is equal to 1, X and X
i
are radically different from each other and in this situation their exchange would introduce a serious error. Therefore, according to the present invention, the filter output is obtained through a superimposition of the effects of several elementary filters whose function is described by equation (1), and such that the output pixel is expressed by the following relationship:
X
out
=&Sgr;X
(
i
)
out
=[&Sgr;K
i
*X+
(1
−K
i
)*
X
i
] (2)
wherein X represents the value relative to the gray level of the pixel being processed; and X
i
represents the value relative to the gray level of a pixel of neighboring the pixel under processing. K
i
provides the measure of similarity between the pixel X and the adjacent pixel X
i
, in practice K
i
represents the value produced by a fuzzy logic processing block according to a membership function corresponding to the sentence “the difference between the pixels is low”. N is the number of neighboring pixels considered.
REFERENCES:
patent: 5680179 (1997-10-01), D'Alto et al.
patent: 5748796 (1998-05-01), Pennino et al.
patent: 5757977 (1998-05-01), Mancuso et al.
patent: 0 415 648 A2 (1990-08-01), None
patent: 0 698 990 A1 (1994-08-01), None
Shen, Q., “Fuzzy Intraframe Smoothing of a Noisy Image,”Electronics Letters, vol. 26, No. 13, Jun. 21, 1990, pp. 908-910.
D'Alto Viviana
Mancuso Massimo
Poluzzi Rinaldo
Sirtori Daniele
Allen Dyer Doppelt Milbrath & Gilchrist, P.A.
Brinich Stephen
Galanthay Theodore E.
Lee Thomas D.
SGS--Thomson Microelectronics S.r.l.
LandOfFree
Adaptive intrafield reducing of Gaussian noise by fuzzy... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Adaptive intrafield reducing of Gaussian noise by fuzzy..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Adaptive intrafield reducing of Gaussian noise by fuzzy... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2482894