An anisotropic diffusion-based defect detection for low-contrast glass substrates

Abstract

In this paper, we propose an anisotropic diffusion scheme to detect defects in low-contrast surface images and, especially, aim at glass substrates used in TFT-LCDs (Thin Film Transistor-Liquid Crystal Displays). In a sensed image of glass substrate, the gray levels of defects and background are hardly distinguishable and result in a low-contrast image. Therefore, thresholding and edge detection techniques cannot be applied to detect subtle defects in the glass substrates surface. Although the traditional diffusion model can effectively smooth noise and irregularity of a faultless background in an image, it can only passively stop the diffusion process to preserve the original low-contrast gray values of defect edges. The proposed diffusion method in this paper can simultaneously carry out the smoothing and sharpening operations so that a simple thresholding can be used to segment the intensified defects in the resulting image. The method adaptively triggers the smoothing process in faultless areas to make the background uniform, and performs the sharpening process in defective areas to enhance anomalies. Experimental results from a number of glass substrate samples including backlight panels and LCD glass substrates have shown the efficacy of the proposed diffusion scheme in low-contrast surface inspection.

Introduction

Surface inspection is an important part of quality control in manufacturing. The manual activity of inspection can be subjective and highly dependent on the experiences of human inspectors. In recent years, image analysis techniques have been increasingly used in industry for surface defect inspection, in which one has to detect small defects that appear as local anomalies in material surfaces. In this paper, we consider the task of automated visual inspection in low-contrast surfaces, and especially focus on the glass substrates used for Thin Film Transistor-Liquid Crystal Displays (TFT-LCDs). The inspection of defects in such panel surfaces ensures the display quality and improves the yield in LCD manufacturing. In the sensed image of a glass substrate, the gray levels of defects and background are hardly distinguishable and result in a low-contrast image. Therefore, simple surface inspection methods such as thresholding and edge detection are difficult to detect subtle defects in low-contrast glass substrate images.

Many defect detection systems aim at uniform surface images such as glass panels [1], sheet steel [2], aluminum strips [3] and web materials [4] using simple thresholding or edge detection techniques. Defects in these uniform images can be easily detected because commonly used measures usually have very distinct values. The surfaces of glass substrates are also a class of uniform images, but with low-contrast intensities. The main low-contrast glass substrates studied in this paper include backlight panels and LCD glass substrates. Fig. 1 presents two types of such glass substrates used for LCD modules. Fig. 1(a1) shows a faultless backlight panel surface, and Fig. 1(b1) shows a defective version of the panel. Figs. 1(c1) and (d1), respectively, present a faultless and a defective LCD panel surfaces. It can be seen from Figs. 1(b1) and (d1) that the defects are difficult to be found in the uniform background with low-contrast intensities. In order to visualize the subtle defects, gray values of glass substrate images are stretched between 0 and 255 for an 8-bit display. Figs. 1(a2)–(d2) show the contrast-stretched images of Figs. 1(a1)–(d1), respectively. The stretched glass substrate images of Figs. 1(b2) and (d2) show the defects clearly, but they also present the background texture and non-uniform illumination. Hence, to detect defects in such stretched images, we may need complicated texture analysis techniques rather than the simple thresholding method. Figs. 1(a3)–(d3) illustrate the gradient images of Figs. 1(a1)–(d1), respectively. These resulting images reveal that the characteristic of a low-contrast surface image invalidates the use of gradient magnitude to identify local anomalies.

In low-contrast surface images, a local defect has a smooth change of brightness from its neighboring region and, therefore, provides no clear edges to apply the gradient-based methods for defect detection. The non-uniform intensity of a faultless region and the low-contrast intensity of a defective region also deter the use of simple thresholding methods. It is extremely difficult to reliably identify small defects in low-contrast surface images without false detection of noise. Little research has been done on defect detection in low-contrast images. Ngan et al. [5] developed an automated vision system for patterned fabrics and repetitive patterned textures. Their system combined wavelet transform and golden image subtraction to detect small-size and low-contrast defects. The method requires a golden image for reference, so the detection performance is affected by environmental changes. Lee and Yoo [6] presented a complicated data fitting approach for detecting regional defects of brightness unevenness in LCD panel surfaces. They first estimated the background surface of an inspection image using a low-order polynomial data fitting. Subtraction of the estimated background surface from the original image was then applied to find the threshold for binary segmentation. The resulting image was then post-processed by median filtering, morphological closing and opening to remove noise and refine the segmentation. The proposed method worked successfully to detect regional defects in low-contrast, non-textured TFT-LCD surface images. However, it is very computationally intensive because the background surface must be estimated recursively by eliminating one pixel at a time throughout the entire inspection image.

In this paper, we propose an anisotropic diffusion scheme to tackle the problem of defect inspection in low-contrast glass substrate images. Anisotropic diffusion was first proposed by Perona and Malik [7] for scale-space description of images and edge detection. It has been widely used as an adaptive edge-preserving smoothing technique for edge detection [8], [9], image restoration [10], [11], image smoothing [12], [13], image segmentation [14], [15] and texture segmentation [16].

The anisotropic diffusion approach is basically a modification of the linear diffusion (or heat equation), and the continuous anisotropic diffusion is given by

$$\frac{\mathrm{\partial }{\mathit{I}}_t(x\text{,}y)}{\mathrm{\partial }t}=\text{div}[c_t(x\text{,}y)·\nabla {\mathit{I}}_t(x\text{,}y)]$$

where I_t(x, y) refers to the image at time t, div the divergence operator, ∇I_t(x, y) the gradient of the image, and c_t(x, y) the diffusion coefficient. If c_t(x, y) is a constant, Eq. (1) is reduced to the isotropic diffusion equation. It is then equivalent to convolving with a Gaussian function. The idea of anisotropic diffusion is to adaptively choose c_t such that intra-regions become smooth while edges of inter-regions are preserved. The diffusion coefficient c_t is generally selected to be a non-negative function of gradient magnitude so that small variations in intensity such as noise or shading can be well smoothed, and edges with large intensity transition are retained.

You et al. [17] gave an in-depth analysis of the behavior of the Perona–Malik anisotropic diffusion model (P–M model) by considering the anisotropic diffusion as the steepest descent method for solving an energy minimization problem. Barash [18] addressed the fundamental relationship between anisotropic diffusion and adaptive smoothing. He showed that an iteration of adaptive smoothing

$${\mathit{I}}_{\mathit{t}+\mathit{1}}(x\text{,}y)=\frac{{\sum }_i{\sum }_j{\mathit{I}}_{\mathit{t}}(x+i\text{,}y+j)w_t(x+i\text{,}y+j)}{{\sum }_i{\sum }_j{w}_t(x+i\text{,}y+j)}$$

is an implementation of the discrete version of the anisotropic diffusion equation if the weight w_t in Eq. (2) is taken as the same of the diffusion coefficient c_t in Eq. (1). Gilboa et al. [19] proposed a forward and backward (FAB) adaptive diffusion process to enhance edge and smooth noise in the image. The FAB diffusion model involves four ad hoc parameters, of which two critical threshold values of gradient must be manually and carefully chosen for the success of the diffusion result. The smaller threshold value determines the use of a forward function, while the larger threshold value determines the use of a backward function. A discontinuous diffusion function is, therefore, applied since nothing is done for the gradient magnitude between the two threshold values.

The conventional diffusion model can effectively perform adaptive smoothing for intra-regions in an image. However, it can only passively stop the diffusion process to preserve original gray values of edges in inter-regions. For defect detection in a low-contrast image, the conventional diffusion model can only smooth the faultless background, but cannot enhance the low-contrast defects. The diffused result may still be a low-contrast image. In this paper, we propose an improved anisotropic diffusion model that aims to enhance the gray level difference between local anomalies and the background to detect defects in low-contrast glass substrate images. The proposed method automatically activates a smoothing process in faultless regions to make the background uniform, and performs a sharpening process in defective regions to enhance anomalies. The proposed method presents a unified continuous diffusion coefficient function that can adaptively carry out smoothing or sharpening operation with only two parameters for defect detection in low-contrast Images. It can distinctly enhance low-contrast defects and uniformly smooth the background without intensifying textured patterns and uneven illumination so that a simple binary thresholding can be effectively and efficiently applied to segment defects in the diffused image.

This paper is organized as follows. In Section 2, we first review the Perona–Malik anisotropic diffusion equation, and then discuss the improved diffusion model that adaptively performs the smoothing and sharpening operations. Section 3 presents the experimental results from a variety of backlight panel and LCD glass substrate surfaces that contain various defects. The effect of varying diffusion parameter values is also analyzed. Finally, Section 4 gives a brief conclusion of our research.