Gray associated analysis based sub-pixel fringe extracting method

Abstract

The invention includes steps: select the reference template for gray correlation analysis and convert the original pictures to pictures of gray correlation by calculating the two-dimensional gray absolute connection; (2) next process the pictures of gray connection with threshold values; (3) then apply the cubic-polynomial-fitting of the two-dimensional gray absolute connection in a small area on the pictures of the gray connection; (4) obtain the fitting parameter by least square method and (5) use the first and second derivatives of the fitting function on the gradient slope to obtain the edge of sub-pixel. This invention for the first time applies the gray system theory for extracting the edge of sub-pixel, effectively suppressing the interferences from white noise and colored noise, enhancing the precision of the edge extraction, reducing the data sizes of the pictures, thus facilitating the operation of the extraction algorithm. This invention can be widely applied in the field of measuring techniques.

Description

technical field [0001] The invention relates to a sub-pixel edge extraction method, in particular to a sub-pixel edge extraction method based on gray relational analysis. Background technique [0002] The edge of the image is the area where parameters such as image grayscale or color change sharply. As a key technology in the process of visual measurement, edge extraction is the basis of target positioning, measurement and evaluation. [0003] Edge extraction is usually performed by first-order differential edge detection operators such as Canny operator (Canny J.F., "A computational approach to edge detection", IEEE Trans. Pattern Analysis and Machine Intelligence. 1986, 8(6): 679-698.) Extraction, the Canny operator is the best first-order differential operator for step-type edges under the influence of white noise. It can realize single-pixel edge extraction and suppress noise interference through smoothing filtering, but its extraction accuracy is only at the pixel level...

AI Technical Summary

Problems solved by technology

The first-order gray moment method (Elbaumand M., Diament P., "Estimation of image centroid size and orientation with laser radar", Applied Optics., 1977, 6: 2438-2440.) is obtained by calculating the first-order gray moment of the edge Sub-pixel edge, this method is easy to calculate, and the translation and scaling of the image do not affect the gray moment, but the accuracy of edge extraction is affected by the uniformity of the gray distribution

The fitting method (Sun Jixiang et al., "Feature Extraction in Pattern Recognition and Computer Vision Invariants", Beijing: National Defense Industry Press, 2001.) is to perform gray value function fitting in a small image area, and then The sub-pixel edge is obtained by calculating the zero-crossing point of its second-order derivative. This method is greatly affected by noise, and the fitting result is unstable when the image area is small, and the accuracy of edge extraction is low.

Spatial moment method (Wang Jianmin et al., "Study on subpixel subdivision algorithm of spatial moment", Optical Technology, 1999, 4:3-6.) is insensitive to additive noise and multiplicative noise in the image, and has high edge extraction accuracy , but near the intersection of two edges, there is a large principle error in the extraction result

[0005] In addition, there are some patented technologies that also propose pixel edge extraction methods, such as the Chinese patent No. 200410102587.3 "A Quick Extraction Method for Sub-Pixel Level Step Edges", which uses a definite edge model to describe the step edge, and uses non- The linear optimization algorithm obtains the model parameters and the position of the boundary points. The initial value selection of the model parameters is simple, the extraction speed is fast, and the robustness is strong. However, due to the use of a definite model to simulate the edge of the image, its applicability is greatly limited. In addition, In the process of edge extraction, this method only performs function fitting in the row or column direction, and does not fully consider the directionality of the edge; the Chinese patent No. 200510123724.6 "A fast and high-precision extraction method for the center of structured light stripes" In fact, it is a fitting interpolation method. The light stripe direction is obtained by calculating the eigenvalues ​​and eigenvectors of the Hessian matrix, and the sub-pixel coordinates of the center of the light stripe are solved according to the Taylor expansion. This method guarantees the extraction accuracy and robustness. In this way, the amount of calculation is reduced, but the algorithm is complex and only suitable for the extraction of line edges

[0006] To sum up, the existing methods or technologies have certain limitations in terms of noise suppression, extraction accuracy, robustness, and computational load, and cannot solve the problem of sub-pixel edge extraction well.

Images