Image analysis – Image enhancement or restoration – Image filter
Reexamination Certificate
1997-12-12
2001-04-10
Grant, II, Jerome (Department: 2624)
Image analysis
Image enhancement or restoration
Image filter
C382S260000, C382S274000, C382S169000
Reexamination Certificate
active
06215911
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention is directed to a method for correcting the intensity of an image, and in particular to a method for correcting an image produced using an image acquisition system which exhibits a sensitivity characteristic over the image space that is non-uniform and that changes slowly.
2. Description of the Prior Art
European Application 0 238 962 discloses a method for digital image correction wherein non-uniform brightness distribution due to an imaging system is corrected in two steps. First, a gross brightness map is produced, from which correction data are obtained by means of smoothing. For the production of the gross brightness map, two methods are proposed: the summation of all brightness values in predetermined windows, and the use of the highest brightness value for the central image point in a predetermined window.
In many image acquisition methods, there is the problem that during the pickup or exposure process a sensitivity characteristic is present that is non-uniform over the imaging region. For example, the sensitivity can fall off toward the edge of the image, so that occlusions or vignettings occur at the edge of the image. An image acquisition method in which this problem occurs is, for example, nuclear magnetic resonance tomography. As is well known, atomic nuclei are thereby excited in a strong magnetic field by means of radio-frequency irradiation. The RF signal that arises is received with an antenna and is converted into intensity values for an image data matrix composed of individual image points (pixels). For the excitation of the nuclear spins, in most cases what a so-called whole-body antenna is used, which exhibits a relatively good homogeneity in the overall region under examination. So-called local coils or local antennas, which cover only a relatively small region, are often used for the reception of the nuclear magnetic resonance signals that arise. In this way, an improved sensitivity and a better signal
oise ratio are achieved. A special form of such local coils is known as an array coil, which has several receive coils arranged alongside one another, and these being respectively switched to an active receive stage according to the size and shape of the region under examination. An example of such an array coil is shown in FIG.
1
. Several overlapping individual coils
1
are thereby provided, which run via matching circuits
2
to a combination network
3
, where the total received signal is finally acquired.
In contrast to the whole-body antenna described above, local coils exhibit a relatively non-uniform sensitivity distribution, i.e., the sensitivity is high in the center and falls off toward the edges. This causes occlusions in the edge regions of the acquired image, which are significantly disturbing. In an array coil as described above, there is the additional problem that the image intensity changes according to the activation of the individual coils. Moreover, flexible local coils are known that are applied directly to the body of the patient. The sensitivity characteristic is thereby not initially determinable, but only results after application to the body of the patient.
In existing apparatuses, e.g. of the type Magnetom® of the company Siemens AG, for the avoidance of this problem a method is used that is briefly presented below in order to explain the underlying problem to which the present invention is directed. It is thereby assumed that, for MR imaging, signals whose origin is encoded in phase factors are acquired in a known way. These signals are demodulated in phase-sensitive fashion, are digitized, and are entered in a raw data matrix, generally referred to as a “k-space.” By means of two-dimensional or three-dimensional Fourier transformation, from this raw data matrix an image data matrix is obtained that is composed of individual pixels with allocated digital intensity values. Imaging by means of magnetic resonance (MR) is not described in more detail herein, since it is not part of the invention. Rather, reference is made to the book by Morneburg—which gives an overview—entitled: Bildgebung fur die medizinische Diagnostik, 3
rd
edition, 1995, chapter 6.2, “Verfahren zur Ortsauflösung.”
In the known method for determining the coil characteristic, it is obtained from the image itself. The correct assumption is made that the coil sensitivity changes only slowly over the image, and thus is expressed in a low-pass-filtered image. Therefore, a strong low-pass filtering of the original image is first carried out, by conducting a two-dimensional Fourier transformation in the row direction and in the column direction in a two-dimensional image. As is well known, by means of the Fourier transformation the image information is converted into the frequency space. In this frequency space, the center corresponds to low frequencies and the edge region corresponds to high frequencies. In the frequency space, a filter is now applied that extracts the central region by filtering. In the simplest case, all values outside a central region, which for example could comprise only 4×4 points in a matrix of 256×256 points, are set to zero. If a two-dimensional inverse Fourier transformation is now applied to the resulting matrix in the frequency space, a return to the image space results, with the auxiliary image acquired in this way being very strongly low-pass filtered in relation to the original image. This auxiliary image is now regarded as the coil characteristic, and is used for the correction of the original image. For this purpose, the auxiliary image is normalized to 1, and in the original image each pixel is multiplied by the reciprocal value of the allocated pixel in the normed auxiliary image. By this procedure, in low-signal regions of the image a very high multiplier could arise, which would, for example, greatly increase the noise. A cutoff is thus predetermined that places an upper limit on the maximum multiplier. This prevents pure noise regions being increased (boosted) to the level of intensity of the rest of the image.
A flow chart of this known method is shown schematically in FIG.
2
. Given an array coil, three original images BD are employed, the respective observation windows being designated FOV (Field Of View), as is standard. By means of a two-dimensional Fourier transformation (2D-FFT), datasets in the frequency space are obtained (FD). The designation 2D-FFT comes from the fact that a two-dimensional Fourier transformation is conducted using, the fast Fourier transform (FFT) algorithm. A filter is applied to the datasets FD, which weights the center thereof—which corresponds to low frequencies—more strongly than the edge regions, which represent higher frequencies. By means of a further Fourier transformation, low-pass-filtered auxiliary images BD
k
are obtained. Taking into account the above-explained cutoff, which is entered by the user, each pixel from the image dataset BD is now multiplied, so that a corrected image BD
k
is obtained.
The effect of this correction is shown in
FIGS. 3
to
5
. A phantom is thereby imaged that essentially has a rectangular cross-section with two indentations. The continuous line represents the target value of the intensity curve I over the axis x. The sensitivity characteristic of the surface coil used to pick up the signal is shown with thin broken lines in FIG.
3
. The intensity curve
1
, shown with a thick broken line, is thereby obtained in the image. Only one noise signal occurs outside the region of the phantom Ph. If the acquired image is strongly low-pass-filtered, the intensity curve shown in
FIG. 4
with a continuous line is obtained. As explained above, this intensity curve is normed to 1, and from the normed values the reciprocal value is calculated, but only up to a determined cutoff, so that the reciprocal value can never become greater than e.g. 5. The intensity values of the original image (thus, of the intensity curve according to
FIG. 3
, shown with a thick broken line) are now
Goertler Georg
Schnur Guenter
Grant II Jerome
Schiff & Hardin & Waite
Siemens Aktiengesellschaft
Worku Negussie
LandOfFree
Method for correcting the intensity of an image 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 correcting the intensity of an image, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for correcting the intensity of an image will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2437085