Process for cyclic, interactive image analysis, and also computer system and computer program for performing the process

• US 6,885,772 B2
• Filed: 04/06/2001
• Issued: 04/26/2005
• Est. Priority Date: 04/08/2000
• Status: Expired due to Fees

• Current Assignee: Delta Transformers, Inc.
• Original Assignee: Carl Zeiss Vision GMBH (Carl-Zeiss-Stiftung)
• Inventors: DeLong, Wolf-Rüdiger
• Primary Examiner(s): MARIAM, DANIEL G
• Application Number:
  US09/828,031
• Publication Number:
  US 20020012466A1
• Time in Patent Office: 1,481 Days
• Field of Search: 382/129, 382/133, 382/159, 382/173, 382/181, 382/190, 382/199, 382/224, 382/258, 382/276
• US Class Current: 382/224
• CPC Class Codes: G06T 7/00 Image analysis

Abstract

The invention relates to a process for cyclic, interactive image analysis with the steps of selection by the user of image regions of interest in the image, performance of various preprogrammed transformations on the image at all image points, performance of a normalization of the transformation results, division of the whole normalized transformation space into classes with the aid of separation rules which are derived from the transformation result values in the image regions selected by the user, allocation of the classification obtained in the transformation space to the individual image points, and representation of the thus classified image in the image space.

Claims

1. A process for cyclic, interactive image analysis, of an image having individual image points, said image defining an image space, comprising the steps of:

(a) selecting by a user of image regions of interest in said image,

(b) performing preprogrammed transformations on said image at one of all individual image points and a selection of individual image points, said performing generating transformation result values defining a transformation space,

(c) normalizing said transformation result values which provides a normalized transformation space,

(d) dividing said normalized transformation space into classes with the aid of separation rules that are derived from normalized transformation result values in said image regions selected by the user, said dividing resulting in a classification,

(e) allocating the classification obtained in said normalized transformation space to said individual image points resulting in a classified image in said image space, and

(f) presenting said classified image in said image space.

2. The process according to claim 1, wherein said preprogrammed transformations include formation of a mean of at least one of color value or gray-scale value and formation of variance of the color value or gray-scale value in an image point neighborhood.

3. The process according to claim 1, wherein said preprogrammed transformations include a Radon transformation for effective representation of lines and edges.

4. The process according to claim 1, wherein division of said normalized transformation space takes place such that regions in the transformation space are identified in which normalized values of each of the performed transformations correspond to normalized values of the same transformation in said image regions selected by the user.

5. The process according to claim 4, wherein a sigmoid function is used for normalization of said normalized values.

6. The process according to claim 1, wherein the process steps (a)-(f) are repeatedly performed on the same image for one of hierarchical refining or generalization of said classification.

7. The process according to claim 1, wherein the process steps (b)-(f) are repeatedly performed on the same image with at least one of different transformations and parameters of said transformations.

8. The process according to claim 1, wherein sequences of transformations, normalizations and classifications are interactively determined and are stored on an storage medium.

9. The process according to claim 8, wherein said interactively determined sequences are subsequently applied to other images.

10. The process according to claim 1, wherein additional properties of said classification are calculated considering classification properties in image regions selected by the user.

11. The process according to claim 10, wherein said additional properties are color values, image brightness, and at least one of variance and class form.

12. A computer system suitable for performance of a process according to claim 1, having devices for interactive input and selection of image regions.

13. A computer program that is stored in a computer readable medium/memory of a digital computer and having a software code for performance of a process with the steps according to claim 1 when said software code runs on said computer.

14. A computer program that is stored in a computer readable medium/memory of a digital computer, having a software code that enables said computer with the computer program loaded into said memory to perform a process with the steps according to claim 1, wherein said software code runs on said computer.

15. The process according to claim 2, wherein said preprogrammed transformations include a Radon transformation for effective representation of lines and edges.