Method for automatic foreground and background detection in digital radiographic images

• US 5,268,967 A
• Filed: 06/29/1992
• Issued: 12/07/1993
• Est. Priority Date: 06/29/1992
• Status: Expired due to Term

• Current Assignee: CareStream Health Incorporated
• Original Assignee: Eastman Kodak Company
• Inventors: Schaetzing, Ralph, Jang, Ben K.
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): Prikockis, Larry J.
• Application Number:
  US07/906,191
• Time in Patent Office: 526 Days
• Field of Search: 382/6, 382/9, 382/22, 382/37, 382/38, 382/50, 382/54, 358/433, 364/413.13
• US Class Current: 382/132
• CPC Class Codes: G06T 2207/10116 X-ray image
• CPC Class Codes: G06T 2207/30004 Biomedical image processing
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/187 involving region growing; involving region merging; involving connected component labelling
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation

Abstract

A digital image processing method automatically separates (segments) the desired regions in a digital radiographic image (i.e., the body part being imaged) from the undesired regions (background areas around the body part that have received unattenuated radiation or foreground areas that have received very little radiation due to the use of radiation limiting devices). The method includes the steps of edge detection, block generation, block classification, block refinement and bit map generation.

Claims

1. A method for automatically segmenting desired regions in a digital radiographic image signal from undesired regions, comprising the steps of:

providing a digital radiographic image signal including a foreground region, a background region, and an object region;

detecting edges in the image signal based on a morphological edge detector;

decomposing said image signal into a set of nonoverlapping, contiguous blocks of pixel signals;

classifying a block of pixel signals to be homogeneous or nonhomogeneous as a function of its edge density and variance;

applying a set of syntactic rules to detect the nonhomogeneous blocks of pixel signals in the foreground/background, as well as the homogeneous blocks of pixel signals in the object;

refining the blocks consisting of a mixture of object and foreground/background; and

producing a segmented image signal.

2. The method of claim 1 wherein the shape and size of a pixel block can be adjusted according to the size of said image signal.

3. The method of claim 1 wherein said refining step refines a pixel block consisting of a mixture of object and foreground/background pixels, based on its local statistics.

4. The method of claim 1 wherein said set of syntactic rules includes one or more of the following rules:

(a) a rule or rules to detect the foreground/background regions which are nonhomogeneous;

(b) a rule or rules to detect object regions which are nonhomogeneous;

(c) a rule or rules to detect the object regions which are homogeneous;

(d) a rule or rules to detect the foreground/background regions which are homogeneous.

5. The method of claim 1 wherein in said applying syntactic rules step a set of clinically and empirically determined decision rules is used to relate various structures of said image signal, and then to detect the nonhomogeneous blocks of pixel signals in the foreground/background, as well as the homogeneous blocks of pixel signals in the object.

6. The method of claim 1 wherein the pixel block classifying step is done by classifying each pixel block into one of ten possible states, depending on its degree of homogeneity and what it contains.

7. The method of claim 6 wherein said ten possible classification states are:

(a) homogeneous background;

(b) homogeneous foreground;

(c) homogeneous object;

(d) nonhomogeneous background;

(e) nonhomogeneous foreground;

(f) nonhomogeneous object;

(g) mixed;

background+foreground;

(h) mixed;

background+object;

(i) mixed;

foreground+object;

(j) mixed;

background+foreground+object.

8. The method of claim 1, including the step of smoothing said digital radiographic image signal before said edge detecting step.

9. The method of claim 8 wherein said smoothing step is effected by a spatial filtering operation.

10. The method of claim 9 wherein said spatial filtering operation includes averaging each pixel of said digital image with the eight nearest neighboring pixels thereof.

11. The method of claim 1 wherein said morphological edge detector produces an edge strength signal, g(x,y), having the following formula:

space="preserve" listing-type="equation">g(x,y)=min[ƒ

B-ƒ

θ

B,ƒ

⊕

B-ƒ

.multidot.B](x,y)

where ƒ

is the smoothed said image signal and B is used to interact with said image signal and is a structuring element which is similar to a kernal in a convolution operation.

12. The method of claim 11 wherein said structuring element of the morphological edge detector can be decomposed into a number of smaller pixel subsets to improve the implementation efficiency.

13. The method of claim 11 wherein said edge strength signal has high-amplitude in areas of strong edges and low-amplitude in areas of weak or no edges.

14. The method of claim 1 wherein said detecting edges step identifies edges as those pixels where said edge strength signal has amplitudes exceeding an edge strength threshold value.

15. The method of claim 14 wherein said edge strength threshold value is automatically determined by maximizing the product of said edge strength threshold and the number of pixels at that threshold.