Enhanced method for reducing ultrasound speckle noise using wavelet transform

• US 5,619,998 A
• Filed: 03/11/1996
• Issued: 04/15/1997
• Est. Priority Date: 09/23/1994
• Status: Expired due to Term

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Rigby, Kenneth W., Abdel-Malek, Aiman A.
• Primary Examiner(s): JAWORSKI, FRANCIS J
• Application Number:
  US08/613,552
• Time in Patent Office: 400 Days
• Field of Search: 128/660.06, 128/660.07, 128/661.01, 73/626, 382/276, 364/413.25
• US Class Current: 600/437
• CPC Class Codes: G01S 7/52028 using digital techniques
• CPC Class Codes: G01S 7/52046 Techniques for image enhancement involving transmitter or receiver image enhancement by image data processing G06T5/00
• CPC Class Codes: G01S 7/52077 with means for elimination of unwanted signals, e.g. noise or interference

Abstract

Signal-dependent noise in a coherent imaging system signal, such as in medical ultrasound imaging, is reduced by filtering speckle noise using nonlinear adaptive thresholding of received echo wavelet transform coefficients, thereby enhancing the resultant image by improving the signal-to-noise ratio. The method includes the steps of dividing the imaging system signal into a number of subinterval signals of equal length; transforming each subinterval signal using discrete wavelet transformation to provide wavelet transform coefficients for each of a plurality of wavelet scales having different levels of resolution ranging from a finest wavelet scale to a coarsest wavelet scale: deleting all of the wavelet transform coefficients representing the finest wavelet scale: for each wavelet scale other than the finest wavelet scale, identifying for each subinterval signal which of the wavelet transform coefficients are related to noise and which are related to a true signal through the use of adaptive nonlinear thresholding; selecting those wavelet transform coefficients which are identified as being related to a true signal; and setting to zero those wavelet coefficients which are identified as being related to noise: inverse transforming the modified wavelet transform coefficients using an inverse discrete wavelet transformation to provide an enhanced true signal with reduced noise.

Claims

1. An adaptive method for reducing signal-dependent noise mixed with a true signal in a coherent imaging system signal, comprising the steps of:

for sampled amplitudes at each beam angle;

(a) dividing a sampled range amplitude signal into a number of first subinterval signals of equal length;

(b) transforming each of said first subinterval signals using discrete wavelet transformation to provide first wavelet transform coefficients for each of a plurality of first wavelet scales having different levels of resolution ranging from a first finest wavelet scale to a first coarsest wavelet scale;

(c) for each transformed beam subinterval, deleting all of the first wavelet transform coefficients representing said finest first wavelet scale;

(d) for each said first wavelet scale other than said finest first wavelet scale, identifying for each of said first subinterval signals which of said first wavelet transform coefficients are related to noise and which are related to a true signal through use of adaptive nonlinear thresholding;

(e) setting to zero those first wavelet transform coefficients which are identified as being related to noise;

(f) inverse transforming the so modified first wavelet transform coefficients using an inverse discrete wavelet transformation to provide a filtered range amplitude signal;

repeating steps (b)-(f) for each subinterval signal comprising said sampled range amplitude signal; and

for filtered amplitudes at each range;

(g) dividing a sampled beam angle amplitude signal into a number of second subinterval signals of equal length;

(h) transforming each of said second subinterval signals using discrete wavelet transformation to provide second wavelet transform coefficients for each of a plurality of second wavelet scales having different levels of resolution ranging from a finest second wavelet scale to a coarsest second wavelet scale;

(i) for each second wavelet scale, identifying for each of said second subinterval signals which of said second wavelet transform coefficients are related to noise and which are related to a true signal through the use of adaptive nonlinear thresholding;

(j) setting to zero those second wavelet transform coefficients which are identified as being related to noise;

(k) inverse transforming the so modified second wavelet transform coefficients using an inverse discrete wavelet transformation to provide a speckle-reduced beam angle amplitude signal; and

repeating steps (h)-(k) for each subinterval signal comprising said sampled beam angle amplitude signal to provide a speckle-reduced range-beam angle image.

2. The adaptive method as defined in claim 1, wherein said adaptive nonlinear thresholding for each said first wavelet scale comprises the steps of:

within a predefined interval, finding for each said first wavelet scale other than the finest first wavelet scale the maximum value of the absolute value of the first wavelet transform coefficients for each said first subinterval signal;

setting the threshold for each said first wavelet scale equal to the minimum of the maximum values for the corresponding wavelet scale, except for the finest first wavelet scale and for the two coarsest first wavelet scales;

setting the threshold for the two coarsest first wavelet scales of said first wavelet transform equal to zero;

within another predefined interval, finding for each said second wavelet scale the maximum value of the absolute value of the second wavelet transform coefficients for each said second subinterval signal;

setting the threshold for each said second wavelet scale equal to the minimum of the maximum values for the corresponding wavelet scale, except for the two coarsest second wavelet scales; and

setting the threshold for the two coarsest second wavelet scales of said second wavelet transform equal to zero.

3. The adaptive method as defined in claim 2, wherein the first selecting step comprises zeroing out all first wavelet transform coefficients having an absolute value less than said threshold set for each said first wavelet scale and passing unchanged said first wavelet transform coefficients having an absolute value which is not less than said threshold set for each said first wavelet scale, and wherein the second selecting step comprises setting to zero all second wavelet transform coefficients having an absolute value less than said threshold set for each said second wavelet scale and passing unchanged said second wavelet transform coefficients having an absolute value which is not less than said threshold set for each said second wavelet scale.

4. The adaptive method as defined in claim 1, wherein the thresholds and the interval lengths are operator selectable.

5. The adaptive method as defined in claim 1, wherein the thresholds are adaptively determined.

6. A speckle noise filter for an ultrasound imaging system, comprising:

means for dividing a sampled range signal into a number of subinterval signals of equal length for each beam;

means for transforming each of said subinterval signals using discrete wavelet transformation to provide wavelet transform coefficients for each of a plurality of wavelet scales having different levels of resolution ranging from a finest wavelet scale to a coarsest wavelet scale;

means for deleting all of the wavelet transform coefficients representing said finest wavelet scale for each transformed beam subinterval;

adaptive nonlinear thresholding means for identifying which of said wavelet transform coefficients are related to noise and which are related to a true signal for those subinterval signals corresponding to wavelet scales other than said finest wavelet scale;

means for selecting only those wavelet transform coefficients which are identified as being related to a true signal; and

means for inverse transforming said selected wavelet transform coefficients using an inverse discrete wavelet transformation to provide an enhanced true signal with reduced noise.

7. The speckle noise filter as defined in claim 6, wherein said adaptive nonlinear thresholding means comprises:

means for finding for each of said wavelet scales, except for the two coarsest wavelet scales, a maximum value of the absolute value of the wavelet transform coefficients for each of said subinterval signals;

means for setting a threshold for each of said wavelet scales, except for the two coarsest wavelet scales, equal to a minimum of the maximum values for the corresponding wavelet scale; and

means for setting a threshold for each of the two coarsest wavelet scales equal to zero.

8. The speckle noise filter as defined in claim 7, wherein said selecting means comprises means for setting to zero all of said wavelet transform coefficients having an absolute value less than said threshold and passing unchanged said wavelet transform coefficients having an absolute value which is not less than said threshold.

9. An ultrasound imaging system comprising:

an array of transducers;

means for selectively operating said transducers in a transmission mode to produce propagating pulses from electrical signals, said electrical signals being timed to produce transmitted ultrasound beams focused at selected points;

means for selectively operating said transducers in a reception mode to produce electrical detection signals from received ultrasound echoes, said electrical detection signals being delayed and summed to produce sampled range signals which are dominated by the received echoes reflected from respective focal points;

a speckle noise filter for filtering out speckle noise included in said sampled range signals;

a scan converter for scan converting said filtered range signals into X-Y display coordinates; and

a display monitor for displaying a series of data samples produced by said means for selectively operating said transducers in a reception mode and scan convened by said scan converter,

wherein said speckle noise filter comprises;

means for dividing the sampled range signal into a number of subinterval signals of equal length for each of said beams;

means for transforming each of said subinterval signals using discrete wavelet transformation to provide wavelet transform coefficients for each of a plurality of wavelet scales having different levels of resolution ranging from a finest wavelet scale to a coarsest wavelet scale;

means for deleting all of the wavelet transform coefficients representing said finest wavelet scale for each of the transformed beam subintervals;

adaptive nonlinear thresholding means for identifying which of said wavelet transform coefficients are related to noise and which are related to a true signal for those of said subinterval signals corresponding to a wavelet scale other than said finest wavelet scale;

means for setting to zero those wavelet transform coefficients which are identified as being related to noise; and

means for inverse transforming the so modified wavelet transform coefficients using an inverse discrete wavelet transformation to provide an enhanced true signal with reduced noise.

10. The ultrasound imaging system as defined in claim 9, wherein said adaptive nonlinear thresholding means comprises:

means for finding for each of said wavelet scales, except for the two coarsest wavelet scales, a maximum value of the absolute value of the wavelet transform coefficients for each of said subinterval signals;

means for setting a threshold for each of said wavelet scales, except for the two coarsest wavelet scales, equal to a minimum of the maximum values for the corresponding wavelet scale; and

means for setting a threshold for each of the two coarsest wavelet scale equal to zero.

11. The ultrasound imaging system as defined in claim 10, wherein the selecting means comprises means for setting to zero all of the wavelet transform coefficients having an absolute value less than said threshold and passing unchanged the wavelet transform coefficients having an absolute value which is not less than said threshold.