Method of detecting intensity edge paths

• US 4,910,786 A
• Filed: 09/30/1985
• Issued: 03/20/1990
• Est. Priority Date: 09/30/1985
• Status: Expired due to Fees

• Current Assignee: Paul H. Eichel
• Original Assignee: Paul H. Eichel
• Inventors: Eichel, Paul H.
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): Couso, Jose L.
• Application Number:
  US06/782,227
• Time in Patent Office: 1,632 Days
• Field of Search: 382/54, 382/53, 382/41, 382/22, 382/27, 382/34, 382/37, 382/38, 382/39, 358/22, 358/182, 358/183, 358/282, 358/284, 378/901
• US Class Current: 382/199
• CPC Class Codes: G06T 2207/10072 Tomographic images
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06T 7/143 involving probabilistic approaches, e.g. Markov random field [MRF] modelling

Abstract

A method of detecting intensity edge paths in an image produced by a technique which is selected from the group consisting of X-ray, CAT scan, nuclear magnetic resonance, photographic, video recording, and photodetection techniques, includes an initial step of determining the gradient magnitude and direction of the image in an array or lattice of node points using a Gaussian weighted gradient operator. A root node having a high probability of being on an edge path is then selected by selecting a node having a high gradient magnitude. The metric for each adjacent node to the root node is then calculated. The various metrics are stored for each such adjacent node. Using the adjacent node having the highest metric, the metrics for other nodes adjacent thereto are calculated. These steps are repeated and the edge path having the highest metric is thereby determined.

Claims

1. A method of detecting intensity edge paths in an image by machine, said image being produced by a technique which is selected from the group consisting of X-ray, CAT scan, nuclear magnetic resonance, photographic, video recording, and photodetection techniques, comprising the steps of:

(a) determine gradient magnitude and direction of the image in an array of node points using a Gaussian weighted gradient operator,

(b) find a root node having a high probability of being on an edge path by selecting a node having a high gradient magnitude,

(c) calculate a path metric, Γ

(mⁱ,fⁱ), for each adjacent node to the root node which may be on the edge by calculating;

##EQU27## where p₁ (f_ri) is the probability that the pixel value F at the location Γ

i arises from an edge , p₀ (f_ri) is the probability that the pixel value F at the location Γ

i does not arise from an edge, and Pr(s_i|s_i-1) is the transition probability for a new transition based on the Markov process,

(d) store the metrics for each such adjacent node.

(e) using the adjacent node having the highest metric, calculate the metrics for other nodes adjacent thereto, and

(f) repeat steps (d) and (e) until the edge path having the highest metric is determined.

2. The method of claim 1 further comprising the additional steps of finding additional root nodes and performing said steps (b)-(f) with respect thereto.

3. The method of claim 2 in which step (f) comprises the step of continuing until an edge path closes on itself, or crosses a previously determined edge path, or reaches the edge of the image.

4. The method of claim 1 in which said metric is a function of the gradient information and the topology of the path by which the node was reached.