A novel wavelet-based multi-resolution skeleton extraction from a moiré image Chien-Yue Chen 1 *, Wing-Kwong Wong 1 **, Tsung-Lun Lin 1 , Ching-Huang Lin 2 1 Institute of Electronic Engineering, National Yunlin University of Science & Technology, Yunlin, TAIWAN 2 Department of Electrical Engineering, Hwa Hsia Institute of Technology, Taipei, TAIWAN *chencyue@yuntech.edu.tw **wongwk@yuntech.edu.tw Abstract: - The measuring of the surface altitude of a body with shadow moiré results in an image of contour lines. These contour lines are usually thick and can be broken at a number of places. A one-pixel wide skeleton of the image can be extracted by thresholding and thinning, which is a traditional approach in previous research. This study proposes a novel method based on wavelet transformation with multi-resolutions and object point detection. In particular, this method solves the problem of uneven lighting on the surface of the object and generates a clear binary image of unbroken contour lines of single-pixel width. Key-Words: - Shadow moiré, contour recognition, wavelet transform, image processing 1 Introduction Moiré, an image of contour lines, is generated by overlapping two grating with similar periods. These grating must be placed with transparent and opulent contours with equi-distance spacing [1]. Moiré images are common in daily life, e.g., textile and silk with grid patterns [2]. If parallel light rays through reference grating are projected on an object, a group of transformed shadows will be formed on the object’s surface, based on the heights of various locations on the surface. These shadows overlap to form an image of non-overlapped lines called shadow moiré. With the computation according to a formula, each pair of neighboring non-overlapped lines differ by a constant height [3]. In other words, the lines form a contour map. Using the technology of optical moiré, one can measure the altitude contours of the surface of an object. In the last decade, the maturation of this technology has resulted in many industrial and scientific applications. This technology can detect the tiny change, which is difficult to detect visually, of the altitude of the surface of an object and magnify the tiny change to a visible scale. This is done without focusing on any local area so that the change of the entire surface can be viewed globally at the same time [4, 5]. In the 1970’s, shadow moiré was first applied to detecting the tiny change in the altitude of the surface of a human body. The contour lines can show the change of the curvature of a patient’s vertebrate [6]. Another application is the measurement of the change of the undulating surface above a muscle of interest [7]. Moiré shadow can also be used to record the tiny change of the surface radius artery of the right hand of a patient [8]. When a shadow moiré is formed on the surface of an object, a researcher often needs to detect whether a pattern of interest is found in the image of shadow moiré. This task can be automated if the image, which in its raw form usually consists of thick and sometimes overlapping or broken contours, can be simplified to just the skeleton, which should be sufficiently thin, of the image. An ideal skeleton image is a binary image with a black skeleton on a white background. In short, it is important to compute at skeleton from a raw contour map. One common method of deriving a binary image from a gray-level image is called bilevel thresholding, which uses the statistical method of a gray-level histogram. Otsu [9], Kittler and Illingworth [10] use such a method but there are a few problems. One problem is its computational time complexity. Moreover the method does not work well for images where the light intensities are not evenly distributed. This problem originates from the way the image is produced. Usually, the image is obtained with a light source projected on the object at an angle. Some locations of the object get more light and others get less, resulting in an uneven distribution of the gray level in the light parts as well as in the dark parts of the image. One solution WSEAS TRANSACTIONS on BIOLOGY and BIOMEDICINE Chien-Yue Chen, Wing-Kwong Wong, Tsung-Lun Lin, Ching-Huang Lin ISSN: 1109-9518 38 Issue 3, Volume 4, March 2007