Processing of oriented patterns

• US 4,745,550 A
• Filed: 08/16/1985
• Issued: 05/17/1988
• Est. Priority Date: 08/16/1985
• Status: Expired due to Fees

• Current Assignee: Schlumberger Technology Corporation (Schlumberger NV)
• Original Assignee: Schlumberger Technology Corporation (Schlumberger NV)
• Inventors: Witkin, Andrew P., Kass, Michael H.
• Primary Examiner(s): Smith, Jerry
• Assistant Examiner(s): Meyer, Charles B.
• Application Number:
  US06/766,639
• Time in Patent Office: 1,005 Days
• Field of Search: 364/421, 364/420, 364/400, 364/422, 382/16, 382/25
• US Class Current: 702/11
• CPC Class Codes: G01V 1/32 Transforming one recording into another or one representation into another
• CPC Class Codes: G01V 2210/40 Transforming data representation for pre-filtering purposes G01V2210/20
• CPC Class Codes: G06T 7/00 Image analysis

Abstract

Disclosed are a method and a system for processing patterns such as oriented patterns, to derive constituents which are more useful than the original patterns. For example, an oriented pattern such as a seismic section can be decomposed into a flow field, which is an estimate of a moving accretion boundary which formed the subsurface layers, and a residual pattern, which is an estimate of layer properties, such as acoustic velocity. Similarly, an oriented pattern such as a chart displaying changes in borehole resistivity with depth along its vertical direction and changes in resistivity with angle around the borehole along its horizontal dimension, can be decomposed into a flow field and a residual pattern, where the flow field is an estimate of a moving accretion boundary and the residual pattern is an estimate of resistivity changes with depth and angle independent of the accretion boundary. The flow field can be mapped onto regular, e.g. orthogonal or polar, coordinates to thereby straighten the original pattern into a deformed or residual pattern which is more useful with respect to visualizing or further processing a parameter of interest than the original pattern.

Claims

1. A method of processing an oriented pattern comprising the steps of:

finding local estimates of flow direction in the pattern;

deriving a flow field based on said local estimates of flow direction; and

extracting useful features of the pattern by utilizing said flow field.

2. A method as in claim 1 in which the extracting step comprises using said flow field to extract a residual pattern from said oriented pattern and wherein the flow field comprises lines conforming to said local estimates, and said lines are mapped on lines conforming to a regular coordinate in the residual pattern.

3. A method as in claim 1 in which the finding step comprises locally estimating the direction of at least one of the greatest and the least change of a selected parameter of the pattern at each of a number of selected locations in the pattern.

4. A method as in claim 3 in which the deriving step comprises deriving a grid comprising (i) flow lines which are along respective directions locally conforming to the local estimates of direction of flow and (ii) transverse lines which are along respective directions locally conforming to directions transverse to said local estimates of direction, wherein said lines intersect each other at points separated along the respective lines by respective constant arc-lengths.

5. A method as in claim 4 in which the extracting step comprises deforming the pattern by mapping said flow lines and transverse lines onto selected regular coordinates.

6. A method as in claim 5 in which the extracting step further comprises extracting selected features of the deformed image by using at least one of said regular coordinates.

7. A method as in claim 1 in which the deriving step comprises deriving a grid comprising (i) flow lines which are along respective directions locally conforming to the local estimates of direction of flow and (ii) transverse lines which are along respective directions locally conforming to directions transverse to said local estimates of direction, wherein said lines intersect each other at points separated along the respective lines by respective constant arc-lengths.

8. A method as in claim 7 in which the extracting step comprises deforming the pattern by mapping said flow lines and transverse lines onto selected regular coordinates.

9. A method as in claim 8 in which the extracting step further comprises extracting selected features of the deformed image by using at least one of said regular coordinates.

10. A method as in claim 1 in which the extracting step comprises deforming the pattern by mapping said flow field onto a coordinate system comprising regular coordinates.

11. A method as in claim 10 in which the extracting step further comprises extracting selected features of the deformed image by using at least one of said regular coordinates.

12. A method as in claim 1 in which the finding step includes filtering the pattern with an orientation-selective linear filter.

13. A method as in claim 12 in which said filtering comprises using a bandpass filter.

14. A method as in claim 13 in which the filter suppresses low and high spatial frequency characteristics of the pattern.

15. A method as in claim 14 in which the filter has a frequency response F which is consistent with the following relationship, in which r is the distance from the origin of the coordinate system, theta is angle, and sigma 1 and sigma 2 are constants selected such that the ratio of sigma 2 to sigma 1 is in the range of about 2 to about 10:

space="preserve" listing-type="equation">F(r,θ

)=[e.sup.r.spsp.2.sup.σ

.sbsp.1.spsp.2 -e.sup.r.spsp.2.sup.σ

.sbsp.2.spsp.2 ]2π

ir cos (θ

).

16. A method as in claim 1 in which the finding step comprises locally estimating the gradient vector of a selected parameter of the pattern of each of a number of selected locations in the pattern, doubling the local gradient vector angles to map local directions differing by about 180°

into a single local direction, summing the resulting gradient vectors over local weighted neighborhoods in the pattern, halving the angles of the resulting summed vectors to estimate the local direction of greatest variance of the selected parameter, and rotating said direction by about 90°

to estimate the local direction of least variance and thus estimate the local flow direction.

17. A method as in claim 16 in which the deriving step comprises starting at a selected location in the pattern and finding grid lines by taking steps of respective fixed arc-length along flow lines which are locally along the local flow directions and along transverse lines which are locally across the local flow directions.

18. A method as in claim 17 in which the step of finding grid lines comprises using bilinear interpolation from said local flow directions.

19. A method as in claim 18 in which the extracting step comprises mapping said grid lines onto a regular coordinate system to derive a deformed pattern.

20. A method as in claim 19 in which said mapping comprises using bilinear interpolation from the values of one or more selected parameters of respective elements of the original pattern to derive interpolated values for elements of the deformed pattern.

21. A method as in claim 19 in which said regular coordinates are orthogonal coordinates.

22. A method as in claim 1 in which the finding step comprises convolving the pattern I(x,y) with an isotropic bandpass filter H(x,y) to produce a convolved pattern C(x,y).

23. A method as in claim 22 in which the filter is applied to each of a number of selected elements of the pattern which are identified by their respective coordinates (x,y), and the filter is characterized by the following expression for H(x,y), in which r is distance and sigma 2 and sigma 1 are selected constants the ratio of which is in the range of about 2 to about 10:

space="preserve" listing-type="equation">H(x,y)=[σ

.sub.1.sup.-2 e.sup.r.spsp.2.sup./σ

.sbsp.1.spsp.2 -σ

.sub.2.sup.-2 e.sup.r.spsp.2.sup./2σ

.sbsp.2.spsp.2 ].

24. A method as in claim 23 in which the convolved pattern C(x,Y) is locally differentiated by finite differences to find the local change in a selected parameter of the pattern in the x-direction (C_x) and in the y-direction (C_y) at each of a number or selected locations (x,y) in the pattern, deriving local measures J₁ (x,y)=2(C_x)(C_y) and J₂ (x,y)=(C_x)² -(C_y)², convolving each of the local measures with a local weighting function W(x,y) to derive respective local measures J₁ *(x,y) and J₂ *(x,y), and deriving the local angle phi of greatest change of a selected parameter of the pattern in accordance with the following approximate relationship:

space="preserve" listing-type="equation">φ

(x,y)≈

tan.sup.-1 (J.sub.1 *(x,y)/J.sub.2 *(x,y))/2.

25. A method as in claim 24 in which the weighting function conforms to the following relationship, in which r is distance and sigma is a constant selected to correspond to the distance between about 2 and about 64 adjacent elements of the pattern:

space="preserve" listing-type="equation">W(x,y)=[1/[[(2π

).sup.1/2 ]σ

]]e.sup.-r.spsp.2 /2σ

.sup.2.

26. A method as in claim 25 including deriving local measures J₃ (x,y)=[C_L³ (x,y)+C_v² (x,y)]^1/2 and convolving them with a local weighting function W(x,y) to derive local measures J₃ *(x,y) and finding the local coherence X(x,y) of the local flow directions in accordance with the relationship:

space="preserve" listing-type="equation">X(x,y)=(J.sub.1 *(i x,y).sup.2 +J.sub.2 *(x,y).sup.2).sup.1/2 /J.sub.3 *(x,y).

27. A method as in claim 26 including the step of displaying the local flow directions and the local flow coherence.

28. A method as in claim 25 in which the deriving step comprises starting at a selected location in the pattern and finding grid lines by taking steps of respective fixed arc-length along flow lines which are locally along the local flow directions and along transverse lines which are locally across the local flow directions.

29. A method as in claim 28 in which the extracting step comprises mapping said grid lines onto a regular coordinate system to derive a deformed pattern.

30. A method as in claim 29 in which the extracting step comprises using said regular coordinates to estimate changes in a selected parameter of the deformed pattern along at least one of said regular coordinates.

31. A method as in claim 1 including the step of deriving measurement signals related to one or more selected parameters of selected subsurface formations, and deriving said oriented pattern from said measurement signals.

32. A process as in claim 31 in which the measurement signals are derived from a borehole traversing selected subsurface formations.

33. A method of processing an oriented pattern comprising the step of:

processing the pattern to estimate its flow field;

using the flow field to represent the pattern in a regular coordinate system based on but different from that of the flow field; and

further processing the pattern in said regular coordinate system to derive desired features of the pattern.

34. A method as in claim 33 including the step of displaying the pattern in said regular coordinates.

35. A method as in claim 34 in which said regular coordinates are orthogonal coordinates.

36. A method as in claim 35 in which said regular coordinates are polar coordinates.

37. A method of processing an original oriented pattern comprising:

processing the original pattern to transform it into a residual pattern which is in a regular coordinate system conforming to the flow field of the original pattern; and

processing the residual pattern by using at least one coordinate of said regular coordinate system to extract desired features of the pattern.

38. A method of processing an oriented pattern comprising:

finding local estimates of flow direction in the pattern; and

using said local estimates to extract an estimate of a residual pattern corresponding to the original pattern but substantially independent of estimated effects of a flow characterized by said local estimates of flow direction.