Granted Patent B2

US 10,475,215 · App. 15/662,670 · Granted Nov 12, 2019

Inventors: Edward B. Gindele (Rochester, NY); Richard A. Simon (Rochester, NY)

Assignee: Carestream Health, Inc.

G06T11/005A61B6/03A61B6/4085A61B6/487A61B6/5205A61B6/5258A61B6/583G06T5/002G06T5/50G06T11/008G06T2207/10076G06T2207/10081G06T2207/20012G06T2207/20016G06T2211/432

Abstract

Claims (46)

1. A method for multi-dimensional image processing comprising:

obtaining a multi-dimensional image having a plurality of data elements, each data element having a corresponding data value;

forming a reduced noise image by a plurality of process iterations estimating the data value for each of one or more data elements of interest p of the obtained image, after obtaining an initial estimate for the data element of interest p, by:

(i) for each data element k in a group of data elements in the image, calculating a weighting factor for the data element k as a function of the difference between the estimated data value for the data element p and the corresponding data value of the data element k; and

(ii) updating the estimated data value for the data element p according to the combined calculated weighting factors and the data value of the data element k; and

displaying, storing, or transmitting the reduced noise image,

wherein calculating the weighting factor for data element k further comprises calculating a similarity metric by:

defining a first region of data elements that includes the data element of interest p;

defining a corresponding region of data elements that includes the data element k; and

generating the similarity metric according to comparison of data values of data elements in the first region about data element p with data values of corresponding data elements in the second region about data element k, and

wherein the weighting factor is computed as a product of the generated similarity metric and a second value that is a function of the difference between the estimated data value for the multi-dimensional data element p and the corresponding data value of the multi-dimensional data element k.

2. The method of claim 1 wherein generating the similarity metric comprises applying a non-local means filter.

3. The method of claim 1 further comprising including the steps of forming the reduced noise image for portions of a Laplacian pyramid representation of the multi-dimensional image.

4. The method of claim 1 wherein the multi-dimensional image is a 2-D or 3-D radiographic image.

5. The method of claim 4 wherein the radiographic image is a cone-beam computed tomography image.

6. The method of claim 1 wherein the weighting factor is further calculated as a function of the separation distance between the data element k and the data element of interest p.

7. The method of claim 1 further comprising adjusting the weighting factor according to an operator instruction.

8. The method of claim 1 wherein the multi-dimensional image is a fluoroscopy image.

9. The method of claim 1 further comprising reducing the resolution of the multi-dimensional image.

10. The method of claim 1 applied to 2-D projection images prior to 3-D volume reconstruction.

11. The method of claim 10 further applied to the 3-D volume reconstruction.

12. The method of claim 1 further comprising transmitting or storing the reduced noise image data values.

13. A method for multi-dimensional image processing comprising:

obtaining a multi-dimensional image having a plurality of data elements, each data element having a corresponding data value;

forming a reduced noise image by a plurality of process iterations estimating the data value for each of one or more data elements of interest p of the obtained image, after obtaining an initial estimate for the data element of interest p, by:

(i) for each data element k in a group of data elements in the image, calculating a weighting factor for the data element k as a function of the difference between the estimated data value for the data element p and the corresponding data value of the data element k; and

(ii) updating the estimated data value for the data element p according to the combined calculated weighting factors and the data value of the data element k; and

displaying, storing, or transmitting the reduced noise image,

wherein the weighting factor is a first weighting factor and further comprising applying a second weighting factor to the data element of interest by:

i) selecting a local region about the data element of interest and a corresponding region about another data element;

ii) for each local data element in the local region about the data element of interest, calculating a similarity metric based on the difference value between the local data element and a corresponding data element within the corresponding region about the other data element; and

iii) calculating the second weighing factor according to the sum of similarity metric values.

14. A method of noise filtering a multi-dimensional image that includes multiple data elements including the steps of:

a) selecting a data element of interest;

b) calculating a reduced noise value for the data element of interest using a combination of a first and second weighting factor;

c) combining the first and second weighting factors to calculate the reduced noise value for the data element of interest;

d) repeating steps b) and c) for other data elements in the multi-dimensional image to form the reduced noise multi-dimensional image; and

e) storing, transmitting, or displaying the reduced noise multi-dimensional image on a display,

wherein the second weighting factor for each other data element is calculated by:

i) selecting a region about the data element of interest and a corresponding region about the other data element;

ii) for each local data element in the region about the data element of interest, calculating a similarity metric based on the difference value between the local data element and its corresponding local data element in the region within the region about the other data element; and

iii) calculating the second weighing factor as the sum of similarity metric values.

15. The method of claim 14 wherein the first weighting factor for each other data element is calculated by:

i) selecting an estimate data value for the data element of interest;

ii) evaluating a weighing function based on the difference between the other data element value and the estimate data value; and

iii) replacing the estimate data value with the reduced noise value and repeating steps i) and ii) one or more times.

Assignments (9)

Continuity (2)