Image analysis – Image transformation or preprocessing – Transforming each dimension separately
Reexamination Certificate
2000-04-17
2003-09-23
Mehta, Bhavesh M. (Department: 2625)
Image analysis
Image transformation or preprocessing
Transforming each dimension separately
C382S298000, C382S282000, C382S255000, C382S275000, C358S463000, C358S453000, C345S616000, C345S613000, C345S611000
Reexamination Certificate
active
06625329
ABSTRACT:
BACKGROUND OF THE INVENTION
The invention relates to a method for improving the definition of a digitized image whose image elements are stored light-intensity values that are organized in rows and columns as an output matrix A
xy
, with a respective matrix section A
ij
of 2
n
×2
m
elements being expanded, through reflection at two of its adjacent edges, and the corner formed by these edges, to 2
n+1
×2
m+1
elements, then subjected to a Fourier transformation. The resulting transformed matrix, B
ij
, 1≦i≦2
n+1
, 1≦j≦2
m+1
, is converted through elementary multiplication with an image-aberration correction matrix C
ij
(all elements being real) into a corrected, transformed matrix B′
ij
, and is transformed back into an intensity matrix A′
ij
, from which a core section is set into the center of a corrected intensity matrix A′
xy
, corresponding to the position of the matrix section A
ij
; then, one matrix section is processed after another in this way, staggered by one edge length of the core section, until a complete, corrected intensity matrix A′
xy
is obtained.
A method of this type is known from Ohm, J. -R.:Digitale Bildcodierung [Digital Image Encoding], Springer 1995, pp. 33/34, 56/57. This processing of local matrix regions with the inclusion of a reflected environment reduces the occurrence of edge disturbances in the image regions, and effects an accelerated processing of the entire matrix.
Moreover, Wahl, F. M. discloses an image processing in the local frequency range only for locally-invariant systems in Digitale Bildsignalverarbeitung [Image Signal Processing], Springer 1984. For image-recording systems possessing locally-varying properties, such as a resolution capability that changes from location to location, local-frequency distortions of image signals are partially compensated with locally-varying filters. In these systems, the local-frequency transmission behavior is a function of the local coordinates. For each image point, therefore, a set of suitable filter coefficients would have to be determined, but to reduce costs, the locally-varying filter only comprises a few filters arranged by image regions.
It is also known to change the stored light-intensity values of the individual pixels of an image in its contrast by adding a background-brightness value and/or through the spreading or normalization of the value range; these modifications can also be implemented separately for different color-image components that are stored as separate value fields. This improves the visual appearance of an image produced with image data that have been modified in this manner; it is not possible, however, to eliminate blurring of the output image with this procedure.
SUMMARY OF THE INVENTION
It is the object of the invention to provide an effective method of improving the definition of images with regard to image aberrations of a lens.
The solution lies in suitably determining the image-aberration correction matrix according to a respective location-dependent aberration-correction matrices from an MTF, that is, modulation-transfer function correction function to match given MTF values of a lens used to image the digitized image, and selecting these local aberration-correction matrices as a subset of the image-aberration correction matrix, corresponding to the position of the associated matrix section, relative to the position of a center point of an image circle, and the distance and the tangential and radial angles of the edges of the matrix section from this central point.
The dependent claims disclose advantageous embodiments.
If the lens and the scale of the image are known, the tangential and radial frequency spread according to the position in the image circle can be taken from the MTF modulation tables of the lens. The correction factor for the individual frequencies results as the reciprocal value of the so-called modulation degree in the respective image circle, namely in the radial and tangential directions to the circle.
If the lens data or the recording data are unknown, according to an embodiment of the method, individual marked image sections are selected, and their processing parameters are varied until a desired optimum definition is achieved. The entire correction matrix can then be derived through fitting from a parametrized correction function. A tried-and-true function is described below.
The novel method operates in the frequency space, with the frequency information for processing the local values being drawn from the surroundings of a small image section. To assure a processing of the individual image sections that is untainted by marginal conditions, the segment's matrix values are reflected three times in different directions, yielding a completely-symmetrical matrix that is twice as large. The spectrum of a matrix of this type is free from marginal interferences.
The entire digital image is gradually processed in this manner, and the matrix values of the individual sections are reworked by an elementary multiplication with a local section of a correction matrix. A subset of the individual result matrix, as the core section, is then transformed back into a result matrix in the intensity range until the matrix is completely formed from these subsets. The core section used after the inverse transformation is preferably a subset corresponding to (2
n
−k
1
)×(2
m
−k
2
); 0<k
1
, k
2
<2
2n
, 2
m
elements of the original matrix section A
ij
, and, accordingly, its edge lengths are 2
n
−k
1
or 2
m
−k
2
, by which core sections are staggered in their arrangement. The known modifications of the elements can therefore be implemented in the amplitude space as desired.
The actual definition of the contours appears in the frequency space through processing, namely through the increase in the high frequency components to the extent that they decreased at the relevant location in the image.
In a further embodiment of the method, a correction function for the radial and tangential image components is derived from the MTF table of the lens. The associated correction function is then calculated from the radial and tangential correction-function values of the two correction functions through interpolation, with consideration of the respective position of the edge of the matrix subset relative to the image circle.
The processing time for an image is heavily dependent on the size of the selected matrix sections. On the other hand, the larger the section, the greater the attainable improvement in definition, although a considerable enlargement does not yield much improvement. In contrast, if the number of matrix elements is too low, the statistical noise is amplified, so raised “seeds and salt” appear on a matte surface: For example, the grain of a reproduced photo or the sieve structure of a reproduced print disadvantageously appears. An image section of 8×8 or 16×16 elements, with a core of 4×4 or 8×8 sharpened elements ultimately being used, has proven advantageous in practice.
Because the image edge of the selected core regions, which are smaller than the sections, does not extend to the original image edge, the inverse-transformed image-element values remaining there after the core has been extracted can be used to fill the edge region. As an alternative, one or more rows of the original matrix elements can be reflected around the image edge prior to the transformation, which creates a slightly larger output matrix, so the core regions arranged together cover the entire original matrix size, and the reflected edge region is omitted and not filled.
In a modification of the method, in the case of a plurality of image-value matrices that are divided by color, they are separated and, provided that they are known, they are corrected with the correction function specified for the respective color. Moreover, the individual color-image matrices are advantageously converted into a matching format through a radial expa
Donner Wilfried
Wöhler Christian
Browdy and Neimark , P.L.L.C.
Kassa Yosef
Mehta Bhavesh M.
LandOfFree
Method for improving the definition of digital images does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for improving the definition of digital images, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for improving the definition of digital images will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3066958