Automatic analysis of two different images of the same object

• US 5,956,435 A
• Filed: 03/24/1997
• Issued: 09/21/1999
• Est. Priority Date: 04/03/1996
• Status: Expired due to Fees

• Current Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Original Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Inventors: Weese, Jurgen, Buzug, Thorsten
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): SHERALI, ISHRAT I
• Application Number:
  US08/822,817
• Time in Patent Office: 911 Days
• Field of Search: 382/130, 382/107, 382/294-299, 382/215, 382/283, 382/154, 382/168-172, 378/1, 378/98.12, 378/41, 707/101
• US Class Current: 382/283
• CPC Class Codes: G06F 18/22 Matching criteria, e.g. proximity measures
• CPC Class Codes: G06T 2207/10072 Tomographic images
• CPC Class Codes: G06T 2207/10116 X-ray image
• CPC Class Codes: G06T 2207/20021 Dividing image into blocks, subimages or windows
• CPC Class Codes: G06T 2207/30004 Biomedical image processing
• CPC Class Codes: G06T 7/30 Determination of transform parameters for the alignment of images, i.e. image registration

Abstract

The invention relates to an automatic image analysis method in which two images which have been recorded at different instants are analyzed and a transformation function is determined on the basis of the shifts of image sections, which transformation function describes the shift from one image to another. In order to enable exact determination of this shift, corresponding image sections are determined in the two images by histogram analysis.

Claims

1. An automatic image analysis method for analyzing a first image and a second image of the same object, the first and the second images comprising pixels having image values, the method comprising:

defining a plurality of image sections (A₁, A₂, . . . A_n) which subdivide an image region (F) to be processed in the first image,

determining for each image section in the first image an optimum shift (v_x, v_y) from the first image to the second image,

wherein, for a particular image section, the optimum shift is selected to be that particular shift (v_x, v_y) for which a difference histogram associated with the particular image section and with the particular shift has the narrowest frequency maximum or maxima, the optimum shift being limited to be within a search area, and

wherein the difference histogram associated with a particular image section and with a particular shift is determined by (i) shifting the particular image section in the first image by the particular shift from its position in the first image to a shifted image section in the second image, (ii) forming a difference image section having pixels with difference image values determined by subtracting image values of first pixels in the particular image section from image values of corresponding second pixels in the shifted image section, a first pixel in the particular image section in the first image corresponding to a second pixel in the shifted image section in the second image if the positions of the two pixels differ by the particular shift, and (iii) forming the difference histogram as a histogram of the frequencies of the difference image values of pixels in the difference image section, and

determining an affine transformation function (X, Y) from the first image to the second image approximating the optimum shifts (y_x, v_y) determined for the image sections.

2. An automatic image analysis method as claimed in claim 1, wherein the distribution (p(b)) given by the difference histogram is converted into a weighted distribution by application of a weighting function (f(p)), wherein the integral over the weighted distribution is calculated, and wherein the difference histogram having the narrowest frequency maximum or maxima is determined as the difference histogram for which this integral has an extreme value.

3. An automatic image analysis method as claimed in claim 2, wherein the weighting function (f(p)) does not have a point of inflection in the range between 0 and 1.

4. An automatic image analysis method as claimed in claim 3, wherein the function f(p)=-p_b log p_b is used as the weighting function, and the difference histogram for which the integral (h) is a minimum is determined.

5. An automatic image analysis method as claimed in claim 1, wherein the first image is a mask image and the second image is a contrast image, and wherein a difference image (D) is derived from the contrast image (C) and the mask image (M), all pixels (x,y) in the image region (F) of the mask image (M) to be processed being subjected to a transformation in conformity with the determined transformation function, and the difference image (D) being derived by subtraction of image values of the transformed mask image and image values of the non-transformed contrast image.

6. An automatic image analysis method as claimed in claim 1, wherein the image region to be processed can be predetermined by the user as an image window (F) which is adjustable in respect to position and size.

7. An automatic image analysis method as claimed in claim 1, wherein the image region (F) to be processed is subdivided into a number of similar, adjoining initial image sections (A₁, A₂ . . . A_n), wherein the contrasts are determined for the individual initial image sections (A₁, A₂ . . . A_n), and wherein the initial image sections with the highest contrasts are selected as the plurality of defined image sections.

8. An automatic image analysis method as claimed in claim 7, wherein the contrasts for the individual initial image sections (A₁, A₂ . . . A_n) are derived from a histogram of the image values in the relevant initial image section.

9. A device for automatic image analysis comprising:

an image-forming device for forming at least two images of an object to be examined,

means for defining a plurality of image sections (A₁, A₂, . . . A_n) which subdivide an image region (F) to be processed in the first image,

means for determining for each image section in the first image an optimum shift (y_x, v_y) from the first image to the second image,

wherein, for a particular image section, the optimum shift is selected to be that particular shift (y_x, v_y) for which a difference histogram associated with the particular image section and with the particular shift has the narrowest frequency maximum or maxima, the optimum shift being limited to be within a search area, and

wherein the difference histogram associated with a particular image section and with a particular shift is determined by (i) shifting the particular image section in the first image by the particular shift from its position in the first image to a shifted image section in the second image, (ii) forming a difference image section having pixels with difference image values determined by subtracting image values of first pixels in the particular image section from image values of corresponding second pixels in the shifted image section, a first pixel in the particular image section in the first image corresponding to a second pixel in the shifted image section in the second image if the positions of the two pixels differ by the particular shift, and (iii) forming the difference histogram as a histogram of the frequencies of the difference image values of pixels in the difference image section, and

means for determining an affine transformation function (X, Y) from the first image to the second image approximating the optimum shifts (v_x, v_y) determined for the image sections.