A kind of magnetic resonance image (MRI) brightness irregularities correction algorithm
Technical field
The present invention relates to magnetic resonance, especially a kind of magnetic resonance image (MRI) brightness irregularities correction algorithm.
Background technology
Magnetic resonance imaging system, because of reasons such as magnetic field and reception unevenness, the situation of brightness irregularities occurs.The reception unevenness of surface coils of wherein take is topmost reason, brighter the closer to the place of surface coils.The image of brightness irregularities not only can cause details in image to be difficult to identification, affects the diagnosis of doctor's read tablet, and affects quantitative analysis results.
Approximate implementation has pre-treating method, and water mould test unevenness before formal scanning is usingd this image scanned as N Reference Alignment.But the method complicated operation, need the foundation of a large amount of raw data acquisitions as correction image, thereby calculated amount is large; The gap of human body and water mould, cause correction error to occur simultaneously.
Summary of the invention
Technical matters to be solved by this invention is to provide a kind of magnetic resonance image (MRI) brightness irregularities correction algorithm, operates fairly simplely, and method has ubiquity, can easily carry out the inhomogeneous correction of brightness of image and process.
For solving the problems of the technologies described above, technical scheme of the present invention is: a kind of magnetic resonance image (MRI) brightness irregularities correction algorithm comprises the following steps:
(1) image I is carried out to the matrix F that Fourier transform obtains frequency domain;
(2) the frequency domain matrix after the conversion adds Gaussian window h and obtains F ', to retain low frequency signal, filtering high-frequency signal;
(3) F ' carries out inversefouriertransform, obtains the low frequency part I of image I
low
(4) I
lowIn be greater than threshold values T point be the too high point of brightness, proofreaied and correct: I (x, y) '=T*I (x, y)/I
low
(5) image I finally obtained '.
The beneficial effect that the present invention compared with prior art brought is:
1) method is simple, based on the image identification irregularity in brightness, does not need to carry out extra test;
2) have applicability more widely, different system and different coil are generally applicable.
The accompanying drawing explanation
Fig. 1 is original image I.
Fig. 2 is original image Fourier transform results F.
Fig. 3 is Gauss function h.
Fig. 4 is for to add Gaussian window F ' as a result to F.
Fig. 5 is image low frequency signal I
low.
Fig. 6 is the too high zone of brightness.
The image I that Fig. 7 is result '.
Embodiment
Below in conjunction with Figure of description, the invention will be further described.
A kind of magnetic resonance image (MRI) brightness irregularities correction algorithm, the algorithm of degree correction requires to retain the details of image, the reason produced according to unevenness, and magnetic field bump and reception susceptibility are slowly to change, this algorithm is processed image according to these characteristics, specifically comprises the following steps:
(1) as Figure 1-3, image I is carried out to the matrix F that Fourier transform obtains frequency domain;
(2) as shown in Figure 4, the frequency domain matrix after conversion adds Gaussian window h and obtains F ', to retain low frequency signal, filtering high-frequency signal;
σ: be defaulted as 30;
R: point (x, y) is to matrix centre distance;
The Gaussian function used is for normalizing on the function basis in the above in [0,1] scope;
It is the value that each pixel of A is multiplied by the correspondence position window function that image A is added to Gaussian window
(3) as shown in Figure 5, F ' carries out inversefouriertransform, obtains the low frequency part I of image I
low
(4) I as shown in Figure 6,
lowIn be greater than threshold values T point be the too high point of brightness, proofreaied and correct: I (x, y) '=T*I (x, y)/I
low
(5) image I finally obtained as shown in Figure 7, '.
In the present embodiment, threshold value T is a constant of determining according to several brightness disproportionation image measurement results, T=10000 in this example (image maximal value 32767).Think that brightness thought bright higher than the point of T, processed by above-mentioned formula, the point (being Fig. 6 zone on the contrary) that is less than T does not process.
As Ilow (343,6)=17953>10000,
I(343,6)’=T*I(343,6)/Ilow(343,6)
=10000*18764/17953=10452
As shown in Figure 1, left figure is the original image gathered, and right figure is the image after algorithm process of the present invention, and by algorithm process of the present invention, the brightness of image unevenness has had very large improvement.
The inventive method is simple, based on the image identification irregularity in brightness, does not need to carry out extra test; Have applicability more widely, different system and different coil are generally applicable.