Yuji Iwahori

Application of fuzzy theory to writer recognition of Chinese characters

Writer recognition is very important for man-machine interface and security. Because each writer has a particular writing form, it is possible to use handwriting characters for writer recognition But a handwriting character is not stable; it changes time to time. This makes writer recognition difficult. To overcome this difficulty, we apply fuzzy theory, because it can absorb the instability of handwriting characters by membership functions. We propose a new method to recognize waters in a short time with a simple algorithm. Three types of membership functions are obtained from normalized hand-writing characters. Three methods, affine transformation (AT), line density equalization (LDE)[1], and four-parts divided affine transformation (FPDAT), are used for normalization. Two types of similarity evaluation functions-ordinary summation of overlapped membership function (SOM), and summation of overlapped and subtraction of un-overlapped membership function (SOSUM)-are tested for 11 Chinese characters written by 16 persons. After evaluating the recognition ratios, we propose the combination of SOSUM and FPDAT. The recognition ratio by this method is 87.9%.

Reconstructing shape from shading images under point light source illumination

A photometric method called point source illuminating stereo is proposed for determining the 3D shape of an object from multiple shading images under point light source illumination. When the surface is a perfect diffuser with uniform reflectance, an algorithm for the determination of 3D shape with positions is developed by using the method of least squares and relying on the principle of the monocular vision and the inverse square law for illuminance. In the proposed method, the number of the necessary images is four for the general surface, and can be reduced to three for the continuous surface.<<ETX>>

Principal components analysis and neural network implementation of photometric stereo

An implementation of photometric stereo is described in which all directions of illumination are close to the viewing direction. This has practical importance but creates a numerical problem that is ill-conditioned. Ill-conditioning is dealt with in two ways. First, many more than the theoretical minimum number of required images are acquired. Second, principal components analysis (PCA) is used as a linear preprocessing technique to extract a reduced dimensionality subspace to use as input. Overall, the approach is empirical. The ability of a radial basis function (RBF) neural network to do non-parametric functional approximation is exploited. One network maps image irradiance to surface normal. A second network maps surface normal to image irradiance. The two networks are trained using samples from a calibration sphere. Comparison between the actual input and the inversely predicted input is used as a confidence estimate. Results on real data are demonstrated

GPU based extraction of moving objects without shadows under intensity changes

This paper proposes a GPU based algorithm for extracting moving objects in real time. The whole process of the proposed approach is handled on GPU. GPU is used for acceleration and the proposed approach increases processing speed dramatically. The method uses a* component and b* component of CIELAB color space without extracting shadow areas as moving objects. It is robust to intensity changes because an estimated background image is generated and moving objects are extracted using background subtraction of the estimated background image and the observed image. The proposed method reduces the times for transferring calculation results from GPU into CPU and the opposite transfer. Reducing the transfer times contributes to speeding up of the proposed method. Results are demonstrated with experiments on real data.

Optimal Edge Detection under Difficult Imaging Conditions

This paper incrementally extends the energy minimization techniques for image analysis developed by [Koch et al. 1986]. Our application is edge extraction and we use the dual intensity and line processes introduced by [Geman and Geman, 1984]. The approach seeks to minimize a global energy functional that explicitly incorporates image properties to be minimized into weighted terms of the energy functional. Our specific contribution is modifying the weighting of terms in the energy functional that were previously independent of spatial gray level change to explicitly include spatial change in the weighting. We argue that the weighting used in previous implementations resulted in a reduced contribution from the edge components due to a dominance of the spatial intensity difference term as that spatial difference increases in size. Our specific modification compensates for this effect by scaling the edge process weighting factors by the spatial difference value (to the second order), thus, maintaining the same relative effect as the spatial difference increases. We found that the proposed algorithm works significantly better as compared to Koch et al. because of this modification.

On the application of cubic equations of state: Analytical expression for α/Tr and improved liquid density calculations

Abstract An analytical equation has been developed for the quantity α/ T r , in the subcritical region of pure fluids for the van der Waals type cubic equations of state, and expressed as a function of the ratio P r / T r and Ω ac . A new approach using an exact fit of two saturated liquid densities was adopted to determine two of the cubic equation parameters. The calculated densities for pure liquids in subcooled conditions or at saturation as well as for mixtures at saturation conditions were much improved on those obtained from the linear volume-translation technique.

Self-calibration and neural network implementation of photometric stereo

This paper describes a new approach to neural network implementation of photometric stereo for a rotational object with non-uniform reflectance factor Three input images are acquired under different conditions of illumination. One illumination direction is chosen to be aligned with the viewing direction. We require no separate calibration object to estimate the associated reflectance maps. Instead, self-calibration is done using controlled rotation of the target object itself. Self-calibration exploits both geometric and photometric constraints. A radial basis function (RBF) neural network is used for non-parametric functional approximation. The neural network training data are obtained from rotations of the target object. Further, the method makes it possible to determine whether or not a given boundary point lies on an occluding boundary. The approach is empirical without needing a distinct calibration object and without making any specific assumptions about the surface reflectance. Experiments on real data are described.

Automatic Detection of Polyp Using Hessian Filter and HOG Features

Abstract An endoscope is a medical instrument that acquires images inside the human body. This paper proposes a new approach for the automatic detection of polyp regions in an endoscope image using a Hessian Filter and machine learning approaches. The approach improves performance of automatic detection of polyp detection with higher accuracy. The approach uses HOG feature as a local feature since the polyp and non-polyp region often have similar color information. The approach also uses Real Adaboost and Random Forests as classifiers which works effciently even when the dimension of feature vector becomes large. It is suggested that Hessian filter can contribute to reducing the computational time in comparison with the case when only HOG features are used to detect the polyp region. K-means++ is introduced to integrate the detection results in the classification. It is shown that polyp detection with high accuracy is performed in the computer experiments with endoscope images.

Human Head Tracking in Three Dimensional Voxel Space

This paper proposes a new approach to track a human head in 3D voxel space. Information of both color and distance is obtained from multiple stereo cameras and integrated in 3D voxel space. Formulating a likelihood function from voxel location and its color information can achieve stable tracking with particle filtering in 3D voxel space

Extended photometric stereo for an object with unknown reflectance property

Conventional Photometric Stereo can obtain the surface gradients of an object when its reflectance property is known. This paper proposes a new method of accurately obtaining the surface gradients with three light source directions under the assumption that the reflectance property of the surface is unknown. Assuming that the reflectance property is uniform over the entire surface, surface gradients can be obtained by the following two steps: (1) reflectance property parameters and the gradients of the suitable surface elements that characterize the reflectance property by solving the nonlinear simultaneous equations for them, so that a pseudo-reflectance map is obtained; and (2) the gradient of each surface element is determined by using parameters obtained by (1). The reliability for the values of the reflectance property parameters (i. e., those for the pseudo-reflectance map) can be improved by applying the method of selecting the surface elements that are sensitive to the reflectance property; this is called the characterized surface elements selection method. In computer experiment with 8-bits shading resolution using three parallel light sources, the values of the reflectance property parameters can be obtained in optimum, and the accurate surface gradients for a glossy object can be determined. A mean square error in a depth measurement becomes an order of 10−2, confirming the usefulness of the proposed method.