Image analysis – Image enhancement or restoration – Focus measuring or adjusting
Reexamination Certificate
1998-09-14
2001-05-08
Rogers, Scott (Department: 2624)
Image analysis
Image enhancement or restoration
Focus measuring or adjusting
C382S261000, C382S279000, C708S315000
Reexamination Certificate
active
06229928
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to an image processing system which easily removes a blur in image data obtained through an optical instrument such as an optical microscope and produces a restored image with an improved picture quality.
The picture quality of an image obtained by observing an object through an optical instrument, such as an optical microscope, deteriorates because of a blur in the image caused by the instrument, as compared with the original luminance distribution of the object. The technique for obtaining an ideal image by removing a blur caused by an optical instrument from the image using numerical computation is known as the restoration of or deconvolution of an image.
A similar technique to this is for emphasizing an image. These techniques are for improving the contrast of an image. The purpose of image restoration is to reproduce the luminance distribution of the original object accurately, whereas the purpose of image emphasis is to clarify the part to be observed at the sacrifice of accurate reproduction of the luminance distribution.
A method of forming images of a specimen under an optical microscope while changing the depth at regular intervals to produce a three-dimensional image (or a stacked image) is known as optical sectioning.
When a stacked image is produced by optical sectioning, a blur (a point spread function: PSF) in the optical microscope spreads more at each of the images gi-1, gi, and g+1 in the direction of depth (or in the direction of z) than in the horizontal direction (or in the direction of xy) as shown in FIG.
1
. As a result, each of the images gi-1, gi, and gi+1 does not have a cross-sectional image accurately reflecting the luminance distribution of the specimen. For example, in the case of the image gi, a blur leaking from each of the overlying image gi-1 and underlying image gi+1 is superposed on the image gi.
One restoration algorithm for removing a blur from such a stacked image easily is the nearest neighbor algorithm. For the nearest neighbor algorithm, refer to, for example, D. A. Agard, “Optical Sectioning Microscopy: Cellular Architecture in Three Dimensions”, Ann. Rev. Biophys. Bioeng, Vol. 13, pp. 191-219, 1984 and D. A. Agard, et at., “Fluorescence Microscopy in Three Dimensions”, Methods in Cell Biology, Vol. 30, pp. 353-377, 1989.
In the nearest neighbor algorithm, only the effect of each of the image gi-1 just above the target image gi and the image gi+1 just blow the image gi is eliminated and the smaller influence of the other planes is ignored. Several types of nearest neighbor algorithms have been proposed according to the degree of approximation.
In the simplest example, a restored image fi is obtained from the i-th stacked image gi and the overlying image gi-1 and underlying image g+1 using the following equation:
fi=c
2
[gi−c
1(
gi-
1
+gi+
1)*
h
] (1)
where c1 and c2 are parameters for adjusting the removal of a blur, h is the value of the point spread function PSF on the overlying and underlying images gi-1, gi+1 when the center of the point spread function PSF is placed in data on the i-th image data item, and * represents convolution.
In such a restoration algorithm, if the sampling interval in the direction of depth of the stacked image is moderately small, the i-th stacked image gi and the overlying and underlying images gi-1, gi+1 will be almost the same. Therefore, even if the overlying and underlying images gi-1, gi+1 are replaced with the i-th image gi, and the nearest neighbor algorithm for the i-th stacked image gi is applied, a blur introduced from each of the overlying and underlying images gi-1, gi+1 will be removed spuriously. This restoration algorithm is known as the no-neighbor algorithm.
In the no-neighbor algorithm, a restored image fi is obtained using the following equation:
fi=c
2[
gi−
2c1(
gi* h
)] (2)
Because in the no-neighbor algorithm, there is no need of referring to the overlying and underlying images gi-1, gi+1, a sheet of image data which is not a stacked image can be processed.
The value h of the point spread function PSF is generally assigned a theoretical value. Discarding the fractions of small values generally give a matrix ranging from 5×5 to 11×11.
Therefore, the convolution of the value h of the point spread function PSF and the stacked image gi constitute a spatial filtering process using h as a coefficient matrix. Differently from a spatial filtering process serving as emphasis means, the no-neighbor algorithm has the advantage that the size and value of the coefficient matrix is always optimized using the theoretical values of the point spread function PSF.
For a method of finding PSF theoretical values, refer to, for example, Y. Hiraoka, et al., “Determination of three-dimensional imaging properties of a light microscope system (Partial confocal behavior in epifluorescence microscopy)”, Biophysical Journal Vol. 57, p. 325-333, February, 1990.
One example of applying the no-neighbor algorithm is an image processing system in a confocal laser scanning microscope (CLSM), whose configuration is as shown in FIG.
2
.
In
FIG. 2
, a CPU
1
drives a scanning driver
2
to scan a convergent light of the laser light on a specimen. A light-receiving element
3
, such as a photomultiplier, receives the light returned from the specimen through a light-receiving pinhole, photoelectrically converts the light into an image signal, and outputs the signal. The image signal is digitized by an A/D converter
4
. The CPU
1
samples the digitized signal and temporarily stores the sampled signal in a memory
5
.
Next, the CPU
1
reads the image data from the memory
5
, do image calculations using equation (2) to produce a restored image fi, and displays the image fi on a monitor television
6
.
After such processing, a high-contrast image with a similar luminance distribution to that of the specimen is obtained.
As described above, in the no-neighbor algorithm, even if there is only one sheet of image data, spurious three-dimensional restoration can be carried out easily on the basis of the point spread function PSF of the optical instrument. As in other types of restoration, the image data is temporarily stored in the memory
5
. Thereafter, the image data is read from the memory
5
and subjected to image calculations to produce a restored image fi. In view of this, the no-neighbor algorithm cannot be used for real-time observation.
Since the confocal laser scanning microscope has a high resolution in the direction of depth, it is characterized by reproducing the three-dimensional luminance distribution of the specimen faithfully. When the light returned from the specimen is faint, however, the diameter of the pinhole on the reception side has to be made larger to compensate for a deficiency of light.
Because making the diameter of the pinhole larger leads to a decrease in the resolution in the direction of depth, the image becomes brighter but its contrast decreases, resulting in a blurred image.
To bring the blurred image into the form of an image with a luminance distribution approximate to that of the specimen by compensating for a decrease in the resolution through restoration, such as the no-neighbor algorithm, the image data has to be stored temporarily in the memory
5
. Thereafter, the image data has to be read and subjected to image calculations to produce a restored image fi.
There is a known method of emphasizing an image signal to display image data more clearly. The method, however, provides no assurance that the displayed image has a faithful reproduction of the actual luminance distribution.
In the case of wide-field optical microscopes, they have a low resolution in the direction of depth inherently. Therefore, they cannot provide an accurate cross-sectional luminance distribution unless suitable restoration is effected.
BRIEF SUMMARY OF THE INVENTION
Accordingly, it is an object of the prese
Frishauf, Holtz Goodman, Langer & Chick, P.C.
Olympus Optical Co,. Ltd.
Rogers Scott
LandOfFree
Image processing system for removing blur using a spatial... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Image processing system for removing blur using a spatial..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image processing system for removing blur using a spatial... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2534868