In this paper, an image processing method to detect thin stains and particles on a surface image is described. The method aims not only to detect the objects but also to distinguish relatively large stains and relatively small particles. In conventional studies, a Fourier transform image processing method to detect the objects is applied. A frequency-transmitting mask applied on a Fourier transformed plane has been used. It is shown that a specific size ring mask enables the detection of specific sized objects. However, when the method is applied to detect stains, it miss-detects elongated particles and the dense particles beside stains. This is because the mask for detecting stains also transmit those particles' frequency components as those particles images contain a wide range of frequency components. The proposed method, first detects the particle image. Then, the original image is subtracted from the processed image to get a particle suppressed image. By applying a stain detecting mask to this image, it was possible to detect stains without miss-detection.

Our previous study showed that adhesive strength can be expressed as a constant value of the intensity of singular stress field (ISSF) by using two-dimensional modelling. In this paper, by considering the three-dimensional (3D) geometry, the mesh-independent technique proposed in our previous study is applied to evaluate the ISSF distributions along the side of a butt-joint interface. The results show that the critical ISSF distributions are almost the same and are independent of the adhesive thickness. The validity of the 2D modelling is confirmed through the present analysis.