CN101396278A - Method of removing real-time ultrasound pattern speckle noise - Google Patents

Abstract

The invention provides to a method for removing real time ultrasound image speckle noise, which comprises the following steps: A1. The size of a filter window and the centre weighting coefficient of the filter window are determined; A2. The pixel Data of a source image is readin; A3. The typical value and the absolute difference of image elements of the filter window which takes every pixel data as a centre are calculated; A4. The weighting coefficient of every image element in the filter window is calculated; A5. The value filtering calculation is carried out on the pixel data of source image; A6. The calculated data is displayed after being digitally scanned and converted. The method realizes the removal of the speckle noise by a software method and replaces the traditional hardware implementation mode, thus saving the cost; the edge and details of the image are well preserved while the speckle noise is removed; the real time tracking of the image is guaranteed.

Description

A kind of method of removing real-time ultrasound pattern speckle noise

Technical field

The present invention relates to the ultrasonic diagnosis technical field, be specifically related to a kind of method of removing real-time ultrasound pattern speckle noise.

Background technology

In ultra sonic imaging, when the physical dimension of tissue and incident ultrasound waves appearance is near or during less than wavelength, ultrasonic beam generation scattering, the scatter echo that phase place is different is interfered the generation speckle noise mutually, it has reduced the quality of ultrasonoscopy, and normal structure and neoplastic lesion tissue in the lower soft tissue of contrast are difficult for respectively.

To studies show that of the statistical property of speckle noise, common speckle noise is obeyed Rayleigh and is distributed, and its average is directly proportional with standard deviation, and this illustrates speckle noise property taken advantage of.But owing in the ultra sonic imaging process, signal has been carried out logarithmic compression, low-pass filtering and interpolation operation etc., thus having changed the statistical property of primary signal, its average no longer is directly proportional with standard deviation, but is directly proportional with variance.Propose a comparatively complicated speckle noise statistical model again in 1989 according to this characteristic Loupas etc. and established x for not by the signal of sound pollution, n is 1 Gaussian noise for, zero-mean variance separate with x, y is observed by the signal of sound pollution, and then the speckle noise model representation of Loupas is:

$y=x+\sqrt{x}*n---\left(1\right)$

At present, the inhibition method of speckle noise has two classes in the ultrasonoscopy.The first kind is a complex method.These class methods are carried out coherence average with the image of one group of same target that certain mode obtains, to remove speckle noise at random.Concrete mode is included in the different time, with different rate of scanning or from different locus tissue is scanned.The method is comparatively ripe, but implementation procedure is comparatively loaded down with trivial details.

Second class methods are filtering methods.From existing document, the technology of present many removal speckle noises is more famous comprises Lee, Kuan and Frost wave filter during being applied in radar data handles.Meanwhile, not to have proposed yet and be applied in the speckle noise inhibition by other wave filter that the speckle statistical model is derived.Such as average rate ripple device, median filter, geometric filter, wavelet transform filter, form rate ripple device, Wiener filtering, also had a lot of new speckle noise removal methods of utilizing preceding several algorithms and other algorithm in recent years in conjunction with development.These methods have reduced the resolution of image to some extent when suppressing speckle noise, and are difficult to guarantee the real-time of image.

In the present low and middle-grade B ultrasonic, utilize FPGA to adopt frame correlation processing method to reach the purpose of removing speckle noise mostly.The frame relevant treatment be the most common also be that the most ancient a kind of time of eliminating speckle is compound.The frame relevant treatment generally is the single order recursive filtering of interframe, the form that can be expressed as:

y(n)＝αy(n-1)+(1-α)x(n) 0<α<1 (2)

The measured value of x (n) expression present frame line sampled point in the formula (2), the filter value of y (n) expression x (n), and the filter value of the same position sampling point of y (n-1) expression former frame, filter coefficient α generally is optional.α is more near 1, and the bandwidth of recursion filter is narrow more, and the effect that suppresses speckle noise is big more, but the real-time tracking of image is poor more.Filter coefficient α is more near 1, and the bandwidth of recursion filter is narrow more, and the effect that suppresses speckle noise is big more.But in the frame relevant treatment algorithm of our existing B ultrasonic system, α generally is taken as 0.5, increases α again, and it is very poor that the real-time tracking of image just becomes, and sampling FPGA carries out the denoising cost comparatively speaking than higher.

Summary of the invention

The technical problem to be solved in the present invention provides a kind of method of removing real-time ultrasound pattern speckle noise, overcomes the method that prior art is removed real-time ultrasound pattern speckle noise, adopts hardware to realize that the image real-time tracking is poor, the denoising defect of high cost.The method that the present invention removes real-time ultrasound pattern speckle noise realizes with software on PC, has both preserved image detail, guarantees the real-time processing speed of image again.

The present invention solves the problems of the technologies described above the technical scheme that is adopted to be:

A kind of method of removing real-time ultrasound pattern speckle noise comprises step:

A1, determine spectral window size and spectral window center weight coefficient;

A2, read in the source image pixel data;

A3, calculating are average and the absolute difference and the average of the spectral window pictorial element at spectral window center with each pixel data;

The weight coefficient of each pictorial element in A4, the calculation of filtered window;

A5, the source image pixel data are carried out medium filtering calculate;

A6, will carry out showing behind the digital scan conversion through the data of calculating.

The average m of spectral window pictorial element calculates according to following formula:

$m=\frac{1}{{(2K+1)}^2}\underset{i=1}{\overset{2K}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{2K}{\mathrm{\&Sigma;}}}X(i,j);$

Wherein: K is the spectral window size factor, and (i j) is the spectral window pictorial element to X.

The absolute difference of spectral window pictorial element and average are calculated according to following formula:

$\mathrm{\&mu;}=\frac{1}{{(2K+1)}^2}\underset{i=1}{\overset{2K}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{2K}{\mathrm{\&Sigma;}}}|X(i,j)-m|.$

Wherein: K is the spectral window size factor, and m is the average of spectral window pictorial element, and (i j) is the spectral window pictorial element to X.

The weight coefficient of each pictorial element calculates according to following formula in the spectral window:

$W(i,j)=W(K,K)-\frac{{\mathrm{\&alpha;d}}_{i,j}\mathrm{\&mu;}}{m};$

Wherein: (K K) is spectral window center weight coefficient to W, and α is a scale factor, and μ is the absolute difference and the average of spectral window pictorial element, and m is the average of spectral window pictorial element, d _{I, j}(i is j) with point (K, distance K) for the spectral window mid point.

Described apart from d _{I, j}Calculate according to following formula:

d _i，j＝Δx+Δy＝|K-i|+|K-j|；

Wherein: i=0,1,2,,, 2K, j=0,1,2,,, 2K,

K is the spectral window size factor.

Described source image pixel data deposit in the two-dimensional array.

Described scale factor and described spectral window center weight coefficient are made as 20 and 99 respectively.

Described scale factor and described spectral window center weight coefficient are made as 20 and 99 respectively.Beneficial effect of the present invention is: the present invention utilizes software approach to realize the removal of speckle noise, replaces traditional hardware implementation mode, provides cost savings; When removing speckle noise, make edge of image and details be able to good preservation; Guaranteed the real-time tracking of image.

Description of drawings

Fig. 1 is spectral window of the present invention and weight coefficient sketch map thereof;

Fig. 2 is a city block distance sketch map of the present invention;

Fig. 3 is a specific implementation process sketch map of the present invention.

The specific embodiment

With embodiment the present invention is described in further detail with reference to the accompanying drawings below:

As shown in Figure 1,

1. the expression size is the pictorial element sequence X () in the spectral window of (2K+1) * (2K+1);

2. the expression size be the weight coefficient sequence W () of (2K+1) * (2K+1), X (i, weight coefficient j) be W (i, j);

d _{I, j}Expression spectral window mid point (i, j) (we calculate with city block distance as shown in Figure 2 for K, distance K) with point;

α is a scale factor;

The average of m presentation video element sequence X ();

μ be pictorial element sequence X () with the absolute difference of average m and average.

W (i, j), d _{I, j}, m, μ draw by formula (3), (4), (5), (6) respectively:

$W(i,j)=W(K,K)-\frac{{\mathrm{\&alpha;d}}_{i,j}\mathrm{\&mu;}}{m},---\left(3\right)$

d _i，j＝Δx+Δy＝|K-i|+|K-j|， (4)

$m=\frac{1}{{(2K+1)}^2}\underset{i=1}{\overset{2K}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{2K}{\mathrm{\&Sigma;}}}X(i,j),---\left(5\right)$

$\mathrm{\&mu;}=\frac{1}{{(2K+1)}^2}\underset{i=1}{\overset{2K}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{2K}{\mathrm{\&Sigma;}}}|X(i,j)-m|,---\left(6\right)$

Wherein i=0,1,2,,, 2K, j=0,1,2,,, 2K.

3. represent the sequence after the weighting.

4. represent the output YWM that weighted median filtering is calculated:

Wherein, MEDIAN{X ₁, X ₂,,, X _NExpression asks the intermediate value of sequence X.

The size of spectral window has determined to remove the degree of noise, experimental results show that when spectral window to be of a size of 9 * 9, and the ability of promptly removing noise during K=4 is the strongest.(K K) has determined hold capacity to edge of image and details, and (K K) gets 20 and 99 respectively for scale factor α and W for scale factor α and window center weight coefficient W.

As shown in Figure 3, X[i] [j], Data[i] [j], Y[i] [j] same point in the correspondence image all, wherein, i=0,1,,, M-1, j=0,1,,,, N-1.

1. process is represented source images X[M] each pixel data X[i in [N]] [j] deposit array Data[i in] in [j], wherein, i=0,1,,, M-1, j=0,1,,,, N-1;

2. process is represented array Data[M] each element Data[i in [N]] [j] value after calculating through adaptive weighted medium filtering deposits array Y[i in] [j], wherein, i=K, K+1,,, M-K, j=K, K+1,,,, N-K, promptly the point to the border does not deal with;

3. process is represented array Y[M] value in [N] shows through behind the digital scan conversion.

Concrete implementation step of the present invention:

1. two array Data[M that size is M * N of application in internal memory] [N], Y[M] [N] view data after being used for depositing source image data, adaptive weighted medium filtering respectively and calculating; Size is the array d[2K+1 of (2K+1) * (2K+1)] [2K+1] be used for depositing the point in the spectral window and the distance of its central point.

2. according to formula (4) preliminary examination d[2K+1] each element d[i of [2K+1]] [j], wherein i=0,1,,, 2K, j=0,1,,,, 2K; Preliminary examination scale factor α; Spectral window size factor K; Spectral window center weight coefficient W[K] [K].

3. read in source images X, suppose that image laterally has N sampled point, vertically has M sampled point.The point that is in the capable j of i row in the source images is deposited in the capable j row of i of Data, i.e. Data[i] [j]=X[i] [j], wherein, i=0,1,,, M-1, j=0,1,,,, N-1.

4. adaptive weighted medium filtering calculates:

1) with Data[M] each element Data[i in [N]] [j] be the center, uses formula (5) and formula (6) to calculate average m and absolute difference and the average μ of its size for the square spectral window pictorial element of (2K+1) * (2K+1) respectively;

2) calculate with Data[i by formula (3)] [j] be that the center size is each interior element Data[i+m of the square spectral window of (2K+1) * (2K+1)] the weight coefficient W[K+m of [j+n]] [K+n], m=-K wherein ,-K-1,,, K-1, k, n=-K ,-K-1,,, K-1, k;

3) by formula (7) calculating pixel data Data[i] [j] output after calculating through weighted median filtering, and this value deposited in Y[i] in [j].

5. Y[M] data in [N] carry out showing behind the digital scan conversion.

Those skilled in the art do not break away from essence of the present invention and spirit, can there be the various deformation scheme to realize the present invention, the above only is the preferable feasible embodiment of the present invention, be not so limit to interest field of the present invention, the equivalent structure that all utilizations description of the present invention and accompanying drawing content are done changes, and all is contained within the interest field of the present invention.

Claims (7)

1, a kind of method of removing real-time ultrasound pattern speckle noise is characterized in that, comprises step:

A1, determine spectral window size and spectral window center weight coefficient;

A2, read in the source image pixel data;

A3, calculating are average and the absolute difference and the average of the spectral window pictorial element at spectral window center with each pixel data;

The weight coefficient of each pictorial element in A4, the calculation of filtered window;

A5, the source image pixel data are carried out medium filtering calculate;

A6, will carry out showing behind the digital scan conversion through the data of calculating.

2, the method for removal real-time ultrasound pattern speckle noise according to claim 1 is characterized in that: the average m of spectral window pictorial element calculates according to following formula:

$m=\frac{1}{{(2K+1)}^2}\underset{i=1}{\overset{2K}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{2K}{\mathrm{\&Sigma;}}}X(i,j);$

Wherein: K is the spectral window size factor, and (i j) is the spectral window pictorial element to X.

3, the method for removal real-time ultrasound pattern speckle noise according to claim 2 is characterized in that: the absolute difference of spectral window pictorial element and average are calculated according to following formula:

$\mathrm{\&mu;}=\frac{1}{{(2K+1)}^2}\underset{i=1}{\overset{2K}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{2K}{\mathrm{\&Sigma;}}}|X(i,j)-m|.$

Wherein: K is the spectral window size factor, and m is the average of spectral window pictorial element, and (i j) is the spectral window pictorial element to X.

4, the method for removal real-time ultrasound pattern speckle noise according to claim 3 is characterized in that: the weight coefficient of each pictorial element calculates according to following formula in the spectral window:

$W(i,j)=W(K,K)-\frac{{\mathrm{\&alpha;d}}_{i,j}\mathrm{\&mu;}}{m};$

Wherein: (K K) is spectral window center weight coefficient to W, and α is a scale factor, and μ is the absolute difference and the average of spectral window pictorial element, and m is the average of spectral window pictorial element, d _{I, j}(i is j) with point (K, distance K) for the spectral window mid point.

5, the method for removal real-time ultrasound pattern speckle noise according to claim 4 is characterized in that: described apart from d _{I, j}Calculate according to following formula:

d _i，j＝Δx+Δy＝|K-i|+|K-j|；

Wherein: i=0,1,2,,, 2K, j=0,1,2,,, 2K,

K is the spectral window size factor.

6, the method for removal real-time ultrasound pattern speckle noise according to claim 5 is characterized in that: described source image pixel data deposit in the two-dimensional array.

7, the method for removal real-time ultrasound pattern speckle noise according to claim 6 is characterized in that: described scale factor and described spectral window center weight coefficient are made as 20 and 99 respectively.

Images