CN1892696B - Supersonic image edge-sharpening and speck-inhibiting method - Google Patents

Abstract

An ultrasound pattern edge sharpening and spot suppression method is used for ultrasonic scanning displayed picture data optimization processing. It contains reading ultrasound pattern data; in turn based on variance judging whether each pixel point being edge point, adopting different data processing calculating to edge point and non-edge point to obtain said point final data and outputting or saving, priority making directivity filtering to said edge point, then making directivity enhancing to obtain enhanced edge; to said non-edge point, finding out pixel data equalizing value and isotropic enhancing value in one adjacent region using said point as central pixel point; finding out both weighted mean value and used as said non-edge point grayscale final value. The present invention can simultaneously realize video edge sharpening and spot suppression.

Description

Ultrasonoscopy edge sharpening and spot inhibition method

Technical field the present invention relates to ultrasonic technique, and the particularly data processing technique in the ultrasonic imaging especially improves the edge sharpening and the spot inhibition method of ultrasonograph quality.

Background technology Fig. 1 is the ultrasonic image-forming system square frame, this system comprises a ultrasonic imaging apparatus, its principle of work is: under the control of master controller, probe receives the ultrasound wave that reflects from tested body tissue to tested body tissue emission ultrasound wave after certain time-delay; This echoed signal enters beam synthesizer, is finished focusing on time-delay, the summation of weighted sum passage by described beam synthesizer; The output signal of this beam synthesizer is given digital scan convertor (D.S.C) and is finished coordinate transform after device detects after testing, finally is sent to the view data that display shows.Wherein system's ultrasonic waves transmitted is actually adding up continuously of back-scattered signal through the scattering of medium and the echoed signal that reflects to form.Inhomogeneous or when having special microstructure features, the ultrasound wave with certain frequency can't be differentiated when dielectric surface, consequent scattered signal has formed the spot on the image.

Be present in the spot in the ultrasonoscopy, show as the black hole in the image, reduced the resolvability and the continuity of image, had a strong impact on picture quality, edge and details are thickened, increased the difficulty of medical diagnosis and treatment.Existing ultrasonic system can also comprise that for this reason an edge sharpening and spot suppress processing module, under the control of described master controller, the view data of described digital scan convertor output is carried out earlier just being sent to display after edge sharpening and spot suppress to handle.Thereby how research suppresses this spot, keeps simultaneously or strengthens image border and minutia, so that ultrasonoscopy is clearer, contrast is stronger, helps diagnosis better, and is significant.

Solving the intrinsic speckle noise of ultrasonoscopy and the main method of edge fog problem at present has:

1) the self-adaptation spot suppresses filtering (seeing that IEEE Trans.Circuits System the 36th in 1989 seizes (1): 129-135 page or leaf, people's such as Loupas T. " An adaptive weighted median filter for speckle suppression inmedical ultrasonic images ").This method belongs to the unsharp masking filtering method, and it adopts different filter factors to control level and smooth degree according to the difference of spot partial statistics characteristic.

2) line rim detection (see the 7th volume of IEEE Trans.Imag.Proc. (12) in 1998: 1700-1714 page or leaf, people's such as Czerwinski R.N " Line and boundary detection in speckle images ").It is similar to the local linear feature (as edge and some special constructions) of image with the very short line segment of length, thisly comes the method for the large scale linearity of approximate diagram image structures with the part than the linearity of small scale, can keep preferably and the enhancing edge.

3) multiple dimensioned non-linear noise suppression and edge strengthen (see the 17th volume of IEEE Trans.Med.Imag. (4) in 1998: 532-540 page or leaf, people's such as Zhang X. " Speckle reduction and contrast enhancement ofechocardiograms via multi-scale nonlinear processing ").Use the analysis of wavelet multiresolution rate, under different scale, use different threshold value noise restraint methods and come level and smooth spot.Its basic thought is that wavelet field is arrived in image transformation, casts out some yardstick part, carries out inverse wavelet transform again, thereby obtains the noise reduction image.

4) anisotropy differential (see nineteen ninety IEEE Trans.Pattern the 12nd volume of Intel1. of Anal.Machine (7): 629-639 page or leaf, people's such as Perona P. " Scale-space and edge detection using anisotropicdiffusion ").In additive process, for neighborhood territory pixel, according to the difference of its shade of gray value, the power of its smoothing effect is also different, and this have anisotropic disposal route in the effect that has uniqueness aspect noise suppression and the preserving edge.

5) ultrasonoscopy that proposes in patent US 6,208,763 and US 6,592,523 of GE company strengthens and spot inhibition scheme, is on the image segmentation basis, based on image gradient image is divided into structural region and background area; Structural region is done directivity filtering and directivity enhancing, isotropic smothing filtering is made in the background area.

Above-mentioned the deficiencies in the prior art are: method 1)-4) though method difference separately all do not have differentiate between images information, and will be used for the whole zone of image with quadrat method; Though method itself can comprise self-adaptive processing, but still can't accomplish to suppress simultaneously speckle noise and strengthen the edge, also make edge fog when when regular meeting occurs strengthening the edge, also having strengthened speckle noise or having suppressed speckle noise situation.Method 1) especially also depend on spot statistical property in the image, and different institutional frameworks is gone up in medical treatment, has different spot statistical properties.Method 5) figure image intensifying effect is more obvious, but it is based on entire image and handles, and the algorithm more complicated is difficult to come real-time implementation with hardware, is difficult to use in actual applications.

The summary of the invention the technical problem to be solved in the present invention is at above-mentioned the deficiencies in the prior art, and a kind of ultrasound image data disposal route is proposed, can handle with simple algorithm and realize simultaneously in real time edge of image sharpening and spot are suppressed, and be easy to realize with hardware mode.

For solving the problems of the technologies described above, of the present inventionly be contemplated that substantially: adopt a kind of spot to suppress and the edge sharpening method, at first judge edge of image, handle respectively again: do directivity filtering or directivity enhancing for the marginal portion based on edge analysis; For non-fringe region, try to achieve average and isotropy enhancing value in the center pixel vertex neighborhood, ask weighted mean to obtain final result to these two values again.Thereby realize simultaneously edge of image sharpening and spot are suppressed.

As the technical scheme that realizes the present invention's design be, a kind of ultrasonoscopy edge sharpening and spot inhibition method be provided, be used for the optimization process of ultrasonic image-forming system, comprise step the ultrasonic scanning image data:

A. read ultrasound image data;

B. successively the data of each pixel are handled, the final data that this point obtains is exported or preserved;

Especially, described step B comprises based on variance judging whether each described pixel is the process of marginal point:

A. be that central point is got a neighborhood with described pixel;

B. obtain the variance Var in the described neighborhood;

C. judge whether marginal point of this central point according to predefined variance thresholding VarT, specifically be, if satisfy condition

Var＞VarT

Then described pixel is a marginal point;

Thereby system adopts different disposal to calculate to marginal point and non-marginal point.

In the such scheme, described step B judges a pixel, and whether marginal point also carries out based on gradient, specifically: be that central point is got a neighborhood with described pixel earlier; Obtain the gradient G rad of described central point; Judge whether marginal point of this central point according to predefined variance thresholding VarT, gradient thresholding GradT and gray scale thresholding GrayT again, the gray-scale value of promptly establishing described central point is gray, if satisfy condition Grad＞GradT, the then described central point of Var＞VarT, Gray＞GrayT is a marginal point.

In the such scheme, system is travel direction filtering to the processing of each marginal point, the edge that is enhanced.

In the such scheme, system comprises also that to the processing of each marginal point the result to directivity filtering further does the directivity enhancing.

In the such scheme, system to the processing of each non-marginal point is: obtaining with this point is pixel data average in the neighborhood of central pixel point, and isotropy enhancing value, again average and isotropy enhancing value are done the final value that weighted mean obtains described non-marginal point.

Adopt technique scheme, can realize simultaneously edge of image sharpening and spot are suppressed, and advantage such as it is simple to have an algorithm, is easy to the hardware realization, can handle in real time, and applicability is strong.

Description of drawings Fig. 1 is the ultrasonic image-forming system block scheme

Fig. 2 is edge sharpening of the present invention and spot inhibition method process flow diagram

Fig. 3 is the edge decision flow chart

Fig. 4 is a gradient calculation template synoptic diagram

Fig. 5 is a directivity filtering process flow diagram

Fig. 6 is that directivity strengthens process flow diagram

Fig. 7 is that isotropy strengthens process flow diagram

Fig. 8 is the original image synoptic diagram that does not adopt the inventive method

Fig. 9 is the image synoptic diagram after the inventive method edge sharpening and spot inhibition

Below the embodiment, the most preferred embodiment shown in is further set forth the present invention in conjunction with the accompanying drawings.

Ultrasonoscopy edge sharpening of the present invention and spot inhibition method, be used for the optimization process of ultrasonic image-forming system to the ultrasonic scanning image data, system reads ultrasound image data earlier, successively the data of each pixel is handled again, and the final data that this point obtains is exported or preserved.Wherein, the system handles module comprises judging whether each described pixel is the process of marginal point as shown in Figure 2, and adopts different disposal to calculate to marginal point and non-marginal point.

Specifically being, in a width of cloth ultrasonoscopy, for each pixel, is that a neighborhood is got at the center with this point, for example includes, but is not limited in the scope of 5*5 pixel, calculates and judges that whether this put marginal point.Specifically be that the variance Var that obtains earlier in the described neighborhood judges whether marginal point of this central point according to predefined variance thresholding VarT again: satisfy condition

Var＞VarT

Be considered as marginal point, otherwise handle by non-marginal point.

Variance Var in the described neighborhood can calculate according to following formula:

$\mathrm{Var}=\left(\underset{i}{\overset{\mathrm{Winsize}}{\mathrm{\Σ}}}\underset{i}{\overset{\mathrm{WinSize}}{\mathrm{\Σ}}}{(\mathrm{Gray}(i,j)-\mathrm{GrayAvg})}^2\right)/(\mathrm{WinSize}*\mathrm{WinSize})$

Wherein WinSize is the neighborhood size, and GrayAvg is the gray average in the neighborhood, and

$\mathrm{GrayAvg}=\underset{j=1}{\overset{\mathrm{WinSize}}{\mathrm{\Σ}}}\underset{i=1}{\overset{\mathrm{WinSize}}{\mathrm{\Σ}}}\mathrm{Gray}(i,j)/(\mathrm{WinSize}\×\mathrm{WinSize})$

In addition, two of embodiments of the invention can be judged marginal point in conjunction with variance and gradient calculation as shown in Figure 3: calculate based on above-mentioned variance, obtain the gradient G rad of described central pixel point again; Judge whether marginal point of this central point according to predefined variance thresholding VarT, gradient thresholding GradT and gray scale thresholding GrayT again, the gray-scale value of promptly establishing described central point is gray, if satisfy condition Grad＞GradT, the then described pixel of Var＞VarT, Gray＞GrayT is a marginal point.

The gradient G rad of described central pixel point can be: Grad=max (fabs (XGrad), fabs (YGrad))

Wherein Xgrad and Ygrad represent the gradient of x and y direction respectively, can calculate that (as shown in Figure 4, figure a is a template of calculating the x direction gradient, and figure b is a template of calculating the y direction gradient with the SOBEL template.Because of belonging to prior art, do not give unnecessary details at this), the maximal value of getting the absolute value of x and y direction gradient like this is the Grad of this pixel.

Neighborhood size in the foregoing description gets 5, can tell the image border preferably.Neighborhood can be obtained bigger, but can increase calculated amount and memory space; And obtain forr a short time, may influence the judgement of image border.

In the embodiment of the invention, be the edge that travel direction filtering is enhanced to the processing of marginal point.Processing to non-marginal point is, obtaining with this point is pixel data average in the neighborhood (being such as but not limited to 5*5) of central pixel point, obtains isotropy enhancing value simultaneously, at last average and isotropy enhancing value done weighted mean and obtains net result.

The detailed process of described directivity filtering is (is example with the 3*3 neighborhood) as shown in Figure 5: to each central pixel point, in neighborhood, calculate earlier the variance of all directions (is example with 0 degree, 45 degree, 135 degree, 90 degree four directions) respectively, relatively obtain minimum variance again, direction along this minimum variance is done mean filter, promptly calculate 3 gray averages of minimum variance direction, at last this average is composed to central pixel point.Described directivity Filtering Processing can make the fuzzy edge smoother that becomes.

For further strengthening the edge, system comprises also that to the processing of each marginal point the result to above-mentioned directivity filtering does the directivity enhancing, the detailed process that described directivity strengthens is (is example with the 3*3 neighborhood) as shown in Figure 6: the one dimension LAPLACIAN value of obtaining all directions earlier, it is { 1 that one dimension LAPLACIAN template can be set, 2 ,-1}; More described again each one dimension LAPLACIAN value is found out maximum amplitude, is made as MaxLAPLACIAN; Calculate the gray-scale value Sharp after the described central pixel point sharpening at last, for

Sharp＝gray+a×MaxLAPLACIAN

Wherein gray is the gray-scale value through directivity filtering; A is the weighting coefficient of presetting, and corresponding with the strong and weak degree of sharpening, different imaging systems is preset different a values.

The detailed process that described isotropy strengthens is (the 5*5 neighborhood that with the corresponding pixel points is the center is an example) as shown in Figure 7: calculate the gray average Mean of this neighborhood interior pixel and the two-dimentional LAPLACIAN value Lap of this central pixel point earlier; Calculate gray scale sharpened value sharp according to described two-dimentional LAPLACIAN value again, as sharp=Lap+gray; Ask the gray scale final value of the weighted mean value of described average Mean and described sharpened value sharp at last as described central pixel point:

result＝(1-b)×Mean+b×sharp

Wherein Yu She weighting coefficient b is used for controlling the fine and smooth degree of image particle in non-marginal point zone (being the background area).Because of the calculating of described one dimension LAPLACIAN and two-dimentional LAPLACIAN belongs to prior art, do not give unnecessary details at this.

It is as follows that described embodiment tests contrast: with the digital black-and-white B of one (but being not limited to) is example, usually image can be obtained display effect as shown in Figure 8, the output of this view data after edge sharpening of the present invention and spot inhibition method processing shows as Fig. 9, as seen, the latter's image border is clearer obviously, speckle noise also obtains fine inhibition, thereby picture quality obtains fine raising.

Come hard-wired said method with DSP (digital signal processor) programming or PLGA (programmable gate array), CPLD (CPLD) mode, also in protection scope of the present invention.

Claims (9)

1. ultrasonoscopy edge sharpening and spot inhibition method are used for the optimization process of ultrasonic image-forming system to the ultrasonic scanning image data, comprise step:

A. read ultrasound image data;

B. successively the data of each pixel are handled, the final data that this point obtains is exported or preserved;

It is characterized in that described step B comprises based on variance judging whether each described pixel is the process of marginal point:

A. be that central point is got a neighborhood with described pixel;

B. obtain the variance Var in the described neighborhood;

C. judge whether marginal point of this central point according to predefined variance thresholding VarT, specifically be, if satisfy condition Var＞VarT

Then described pixel is a marginal point;

Thereby system adopts different disposal to calculate to marginal point and non-marginal point.

2. ultrasonoscopy edge sharpening according to claim 1 and spot inhibition method is characterized in that, the variance Var in the described neighborhood is:

$\mathrm{Var}=\left(\underset{i}{\overset{\mathrm{Winsize}}{\mathrm{\Σ}}}\underset{i}{\overset{\mathrm{WinSize}}{\mathrm{\Σ}}}{(\mathrm{Gray}(i,j)-\mathrm{GrayAvg})}^2\right)/(\mathrm{WinSize}*\mathrm{WinSize})$

Wherein, WinSize is the neighborhood size, and GrayAvg is the gray average in the neighborhood, and

$\mathrm{GrayAvg}=\underset{j=1}{\overset{\mathrm{WinSize}}{\mathrm{\Σ}}}\underset{i=1}{\overset{\mathrm{WinSize}}{\mathrm{\Σ}}}\mathrm{Gray}(i,j)/(\mathrm{WinSize}\×\mathrm{WinSize}).$

3. ultrasonoscopy edge sharpening according to claim 1 and spot inhibition method is characterized in that,

Described step B judges a pixel, and whether marginal point also carries out based on compute gradient, specifically: be that central point is got a neighborhood with described pixel earlier; Obtain the gradient G rad of described central point; Judge whether marginal point of this central point according to predefined variance thresholding VarT, gradient thresholding GradT and gray scale thresholding GrayT again, the gray-scale value of promptly establishing described central point is gray, if satisfy condition

Grad＞GradT，Var＞VarT，Gray＞GrayT

Then described central point is a marginal point.

4. ultrasonoscopy edge sharpening according to claim 3 and spot inhibition method is characterized in that, the gradient G rad of described central point is:

Grad＝max(fabs(XGrad)，fabs(YGrad))

Wherein, XGrad and YGrad represent the gradient of x and y direction respectively.

5. ultrasonoscopy edge sharpening according to claim 1 and spot inhibition method is characterized in that:

System is travel direction filtering to the processing of each marginal point, the edge that is enhanced.

6. ultrasonoscopy edge sharpening according to claim 5 and spot inhibition method is characterized in that, described system comprises step to the described directivity filtering of each described marginal point:

A. be that central point is got a neighborhood with described marginal point;

B. in this neighborhood, calculate the variance of all directions respectively;

C. relatively obtain minimum variance;

D. calculate the gray average of this above neighborhood interior pixel point of minimum variance direction, and this average is composed to described marginal point.

7. ultrasonoscopy edge sharpening according to claim 6 and spot inhibition method is characterized in that, described system comprises also that to the processing of each described marginal point the result to described directivity filtering further does the directivity enhancing, comprises step:

A., one dimension LAPLACIAN template is set obtains the one dimension LAPLACIAN value of all directions in the described neighborhood;

B. the one dimension LAPLACIAN value of more described all directions is found out maximum amplitude, is made as MaxLAPLACIAN; Calculate the gray-scale value Sharp after the described marginal point sharpening

Sharp＝gray+a×MaxLAPLACIAN

Wherein, gray is the gray-scale value of described marginal point through directivity filtering; A is the weighting coefficient of presetting.

8. ultrasonoscopy edge sharpening according to claim 1 and spot inhibition method is characterized in that,

System to the processing of each non-marginal point is: obtaining with this point is pixel data average in the neighborhood of central pixel point, and isotropy enhancing value, again average and isotropy enhancing value is done the final value that weighted mean obtains described non-marginal point.

9. ultrasonoscopy edge sharpening according to claim 8 and spot inhibition method is characterized in that,

The detailed process that described isotropy strengthens comprises: calculate the gray average Mean of described neighborhood interior pixel and the two-dimentional LAPLACIAN value Lap of described central pixel point earlier; Calculate gray scale sharpened value sharp according to described two-dimentional LAPLACIAN value again

sharp＝Lap+gray；

Ask the weighted mean value of described average Mean and described sharpened value sharp at last

result＝(1-b)×Mean+b×sharp

Gray scale final value as described central pixel point; Wherein b is default weighting coefficient.

Images