B-Spline Curve Fitting and the Feature Point Extraction of Fiber Image

Abstract

Fiber images taken by optical microscope tend to have poor focus, uneven brightness or blurred edge. Using mathematical morphology method, such as open or close operation to preprocess the fiber image, which is likely to lead to distortion of the edge of the fiber image and affect the subsequent separation operation. B-spline curve is used to fit the background of the image, then the fitting background is subtracted from the original image. Next, the complex domain nonlinear anisotropic diffusion process is used to denoise and enhance the image edge. Finally, Binarized the image and filling area to get a clear picture of the target fiber image. Introduction At present, the method of computer automatic recognition technology of fiber image has achieved more attention by domestic and foreign researchers[1-4], several features for evaluation have also been put forward. However, the available technology of intelligent fiber identification depends on a number of extraction of feature indicators, such as longitudinal and cross-sectional morphological characteristics of cotton fibers. These indicators are often based on precise measurements of the individual fibers, but the actual test of the required sample image usually includes a plurality of fibers. So how to separate a plurality of overlapping target is the first problem to solve. Fiber images taken by optical microscope tends to have poor focus, uneven brightness or blurred edge. Using mathematical morphology method, such as open or close operation to preprocess the fiber image, which is likely to lead to distortion of the edge of the fiber image and affect the subsequent separation operation. Medical image processing[5] can adopt methods as image contour pits, watershed algorithm, and morphology and corrosion algorithm for segmentation of overlapping cells or drugs. Their shapes are relatively regular but different from those of the fibers. For a fiber image, the phenomenon of cross, torsion or adhesion between the fibers makes the separation very difficult. In this paper, the B-spline curve fitting and complex domain nonlinear anisotropic diffusion process are used to deal with the edge of fiber image. B-spline curve is used to fit the background of the image, and the fitting background is subtracted from the original image. Next, the complex domain nonlinear anisotropic diffusion process is used to denoise and enhance the image edge. Finally, the binarization and region filling processing is used to get the clear fiber object. Algorithm Description The B-spline curve gray fitting of fiber image The B-spline curve fitting method is used to get the fitting background of image, then the fitting background is subtracted from the original fiber image to obtain the target image. This process is able to remove the image background and improve the uneven illumination impact of image. The algorithm uses a bi-cubic B-spline curve to fit the background, restricted by the least squares. The gray value of fiber image is defined as F(xi, xj), where 1 M, n > N. For a given range: Xu : 0 = u0 < ... <um=1, Yv: 0 = v0 < ... <vn=1, the Xu and Yv respectively have (m + 1) and (n + 1) cubic B-spline function which are recorded as Bi(u) and Bj(v), where i = 0, 1, ...m, and j = 0, 1, ...n. 4th International Conference on Computer, Mechatronics, Control and Electronic Engineering (ICCMCEE 2015) © 2015. The authors Published by Atlantis Press 794 Accordingly, the background of the image can be expressed as follows: ∑∑ = = = m