Du-Ming Tsai

Independent Component Analysis-Based Background Subtraction for Indoor Surveillance

In video surveillance, detection of moving objects from an image sequence is very important for target tracking, activity recognition, and behavior understanding. Background subtraction is a very popular approach for foreground segmentation in a still scene image. In order to compensate for illumination changes, a background model updating process is generally adopted, and leads to extra computation time. In this paper, we propose a fast background subtraction scheme using independent component analysis (ICA) and, particularly, aims at indoor surveillance for possible applications in home-care and health-care monitoring, where moving and motionless persons must be reliably detected. The proposed method is as computationally fast as the simple image difference method, and yet is highly tolerable to changes in room lighting. The proposed background subtraction scheme involves two stages, one for training and the other for detection. In the training stage, an ICA model that directly measures the statistical independency based on the estimations of joint and marginal probability density functions from relative frequency distributions is first proposed. The proposed ICA model can well separate two highly-correlated images. In the detection stage, the trained de-mixing vector is used to separate the foreground in a scene image with respect to the reference background image. Two sets of indoor examples that involve switching on/off room lights and opening/closing a door are demonstrated in the experiments. The performance of the proposed ICA model for background subtraction is also compared with that of the well-known FastICA algorithm.

Fast normalized cross correlation for defect detection

Normalized cross correlation (NCC) has been used extensively for many machine vision applications, but the traditional normalized correlation operation does not meet speed requirements for time-critical applications. In this paper, we propose a fast NCC computation for defect detection. A sum-table scheme is utilized, which allows the calculations of image mean, image variance and cross-correlation between images to be invariant to the size of template window. For an image of size M × N and a template window of size m × n, the computational complexity of the traditional NCC involves 3 ċ m ċ n ċ M ċ N additions/subtractions and 2 ċ m ċ n ċ M ċ N multiplications. The required numbers of computations of the proposed sum-table scheme can be significantly reduced to only 18 ċ M ċ N additions/subtractions and 2 ċ M ċ N multiplications.

Automated surface inspection for directional textures

Abstract In this paper we present a global approach for the automatic inspection of defects in directionally textured surfaces which arise in textile fabrics and machined surfaces. The proposed method does not rely on local features of textures. It is based on a global image restoration scheme using the Fourier transform. The line patterns of any directional textures in the spatial domain image are removed by detecting the high-energy frequency components in the Fourier domain image using a one-dimensional (1D) Hough transform, setting them to zero, and finally back-transforming to a spatial domain image. In the restored image, the homogeneous line region in the original image will have an approximately uniform gray level, whereas the defective region will be distinctly preserved. A statistical process control scheme is therefore used to set up the control limits for discriminating between defects and homogeneous line patterns. The experiments on a variety of textile fabrics, machined surfaces and natural wood have shown the effectiveness of the proposed method.

A fast thresholding selection procedure for multimodal and unimodal histograms

In this paper, a simple and efficient histogram-based approach is presented for multi-level thresholding. It uses Gaussian kernel smoothing to detect peaks and valleys in a multimodal histogram, and uses a local maximum curvature method to detect points of discontinuity in a unimodal histogram. The computational time will decrease as the desired number of thresholding levels increases. The performance of the proposed algorithm is compared with those of the widely applied between-class variance and entropy methods.

Automated surface inspection for statistical textures

Abstract In this paper we present a global approach for the automatic inspection of defects in randomly textured surfaces which arise in sandpaper, castings, leather, and many industrial materials. The proposed method does not rely on local features of textures. It is based on a global image reconstruction scheme using the Fourier transform (FT). Since a statistical texture has the surface of random pattern, the spread of frequency components in the power spectrum space is isotropic and forms the shape approximate to a circle. By finding an adequate radius in the spectrum space, and setting the frequency components outside the selected circle to zero, we can remove the periodic, repetitive patterns of any statistical textures using the inverse FT. In the restored image, the homogeneous region in the original image will have an approximately uniform gray level, and yet the defective region will be distinctly preserved. This converts the difficult defect detection in textured images into a simple thresholding problem in nontextured images. The experimental results from a variety of real statistical textures have shown the efficacy of the proposed method.

A simulated annealing approach for optimization of multi-pass turning operations

In this paper, an optimization algorithm based on the simulated annealing (SA) algorithm and the Hooke-Jeeves pattern search (PS) is developed for optimization of multi-pass turning operations. The cutting process is divided into multi-pass rough machining and finish machining. Machining parameters are determined to optimize the cutting conditions in the sense of the minimum unit production cost under a set of practical machining constraints. Experimental results indicate that the proposed nonlinear constrained optimization algorithm, named SA/PS, is effective for solving complex machining optimization problems. The SA/PS algorithm can be integrated into a CAPP system for generating optimal machining parameters.

Optimal multi-thresholding using a hybrid optimization approach

The Otsus method has been proven as an efficient method in image segmentation for bi-level thresholding. However, this method is computationally intensive when extended to multi-level thresholding. In this paper, we present a hybrid optimization scheme for multiple thresholding by the criteria of (1) Otsus minimum within-group variance and (2) Gaussian function fitting. Four example images are used to test and illustrate the three different methods: the Otsus method; the NM-PSO-Otsu method, which is the Otsus method with Nelder-Mead simplex search and particle swarm optimization; the NM-PSO-curve method, which is Gaussian curve fitting by Nelder-Mead simplex search and particle swarm optimization. The experimental results show that the NM-PSO-Otsu could expedite the Otsus method efficiently to a great extent in the case of multi-level thresholding, and that the NM-PSO-curve method could provide better effectiveness than the Otsus method in the context of visualization, object size and image contrast.

Boundary-based corner detection using eigenvalues of covariance matrices

Abstract In this paper we present a new measure for corner detection based on the eigenvalues of the covariance matrix of boundary points over a small region of support. It avoids false alarms for superfluous corners on circular arcs. Experimental results have shown that the proposed corner detection methods using curvature measures have good detection and localization for curved objects in different rotations and with varying scale changes.

Rotation-invariant pattern matching using wavelet decomposition

In this paper, we propose a wavelet decomposition approach for rotation-invariant template matching. In the matching process, we first decompose an input image into different multi-resolution levels in the wavelet-transformed domain, and use only the pixels with high wavelet coefficients in the decomposed detail subimage at a lower resolution level to compute the normalized correlation between two compared patterns. To make the matching invariant to rotation, we further use the ring-projection transform, which is invariant to object orientation, to represent an object pattern in the detail subimage. The proposed method significantly reduces the computational burden of the traditional pixel-by-pixel matching. Experimental results on a variety of real images have shown the efficacy of the proposed method.

Automatic band selection for wavelet reconstruction in the application of defect detection

Abstract In this paper, we present a multiresolution approach for the inspection of local defects embedded in homogeneously textured surfaces. It is based on an efficient image restoration scheme using the wavelet transforms. By properly selecting the smooth subimage or the combination of detail subimages at different resolution levels for image reconstruction, the global repetitive texture pattern can be effectively removed and only local anomalies are preserved in the restored image. A wavelet band selection procedure is developed to automatically determine the best reconstruction parameters based on the energy distribution of wavelet coefficients. Experimental results show that the decomposed subimages and the number of resolution levels determined by the automatic band selection scheme are similar to the manual selection results, and the defects in a variety of real textures including machined surfaces, natural wood, sandpaper and textile fabrics are well detected.