Shozo Kondo

DETECTION OF ROADS FROM SATELLITE IMAGES USING OPTIMAL SEARCH

Since networks of main roads are basic information for the classification of use of the earth surface, the automatic detection of roads from satellite images is a very important issue. In this paper, a new detection theory is proposed which can overcome drawbacks of current theories and detect plural roads in an image with high speed and high precision. Firstly, binary images representing edges are used to evaluate the possibility for a road to pass on a given pixel. An 8-directions-filter, a clearing filter and a parallel-edge-detection filter are proposed which can bring insufficient local information to each other to obtain global information enough to detect a road and by which the possibility of a road-passing on the pixel can be effectively evaluated. Secondly, by using the Hough transform and the optimal search method it is possible to detect a complete road. This detection theory does not depend on the size of image and can detect all the roads in an image including intersecting and T-type roads.

Detection of roads from satellite image using optimal search

The automatic detection of roads from satellite images is a very important problem. In this paper a new method is proposed which can overcome drawbacks of current road detection theories and can detect multiple roads in an image. Firstly binary images representing edges are used to evaluate the possibility for a road to pass on a given pixel. An 8-directions filter, a clearing filter and a parallel-edge-detection filter are proposed which can bring insufficient local information to each other to obtain enough global information to detect a road and by which the possibility of a road-passing on the pixel can be effectively evaluated. Secondly, by using the Hough transform and the optimal search method a complete road can be detected. This detection theory does not depend on the size of an image and can detect multiple roads in an image including intersecting roads and T-type roads.

Off-line handwritten Thai characters from word script

This paper proposes a recognition method of off-line handwritten Thai characters from word scripts. Firstly a new method on line level separation of Thai words are proposed. Loop structure is used to classify 4 groups of characters. Secondly by using topological properties of strokes and other Thai characters structural features decision trees are constructed. Finally a recognition experiment is presented in which 100 copies of handwritten Thai words written by 10 persons are tested. Recognition rate is 99.0% and recognition time is 0.5 second per character.

A New Content-based Image Retrieval Using Color Correlogram and Inner Product Metric

The color correlogram is a simple statistical descriptor of a color image that has been widely used for content-based image retrieval (CBIR) systems. To measure similarity between two images using the correlogram, the traditional approaches use the relative distance. In this paper, to improve performance of the CBIR systems, the inner product metric is used to measure similarity of images instead of the relative distance. Results of experiments proved that the CBIR using the inner product metric has better performance than the one using the relative distance.

Detection of roads from satellite image using the optimal search

Since networks of main roads are basic information for classification of the use of the Earths surface, automatic detection of roads from satellite images is a very important problem. In this paper, a new detection theory is proposed which can overcome drawbacks of current theories and detect plural roads in an image with high speed and high precision. Firstly, a binary image representing edges of a given image is used to evaluate the possibility for a road to pass on each edge pixel. We propose an 8-direction filter, a clearing filter, and a parallel-edge-detection filter, which can compile insufficient local information to obtain global information enough to detect a road. After these filters, the possibility of a road passing on a edge pixel can be effectively evaluated. Secondly, by using the Hough transform and the optimal search method it is possible to detect a complete road. This detection theory does not depend on the size of image and can detect almost all the main roads in an image including intersecting and T-type roads.

A New Similarity Measure for Content-Based Image Retrieval Using the Multidimensional Generalization of the Wald-Wolfowitz Runs Test

This paper proposes a new similarity measure for the content-based image retrieval (CBIR) systems. The similarity measure is based on the multidimensional generalization of the Wald-Wolfowitz (MWW) runs test and the k-means clustering algorithm. The performance comparisons between the proposed method and the current CBIR method based on MWW runs test were performed, and it can be seen that the proposed methods outperform the current method in the sense that the proposed method provides higher performance than the current method for the same computational time.

An Improved Wavelet-based Watermarking Method Using the Mathematical Morphology

In this paper, we propose an improved wavelet- based watermarking method using the mathematical morphology. The wavelet coefficients of host image are firstly selected as the significant coefficients by a statistical predetermined threshold. The mathematical operations are then performed to tame some small perturbations on the selected coefficients caused by the manipulations or attacks. Moreover, the pixel-wise masking model which exploits the characteristics of human visual system is employed for improving watermark invisibility and robustness. The experimental results verify that our method is more robust and effective than the existing method.

A theory of image restoration for linear spatial degradation using multiresolution analysis

In this theory it is assumed that an operator representing linear degradation is unknown but a pair of original images and its degraded image is given. The two images are projected onto several image spaces whose spatial frequencies are included in specified bands using the MRA. By comparison between the two projected images in the image spaces, demodulation coefficients are obtained which are used to restore the degraded image to the original one. In the process of obtaining the coefficients a new image distance is used which is a comparable measure that would correspond to the way the human visual system assesses. The obtained demodulation coefficients can be used to restore other degraded images. Examples of image restorations support the effectiveness of the proposed method.

Robust 2D Fisher Barycenter Contour

In this paper, the algorithm for 2D shape matching and retrieval is developed by using Fisher Barycenter Contour (FBcC). First, the shape is represented into 3D format using the signed enclosed area at each scale level of Barycenter Contour (BcC). Because of high dimension of the feature representation, the Eigen Barycenter Contour (EBcC) is applied for dimensionality reduction. Then, the Fisher Barycenter Contour (FBcC) is used for making discrimination. Finally, the similarity is measured using the normalized cross correlation. The experimentation is tested on MPEG-7 contour shape database CE-1 part B of 1400 image shapes. The experimental results illustrate that our approach gives very high retrieval efficiency (or Bulls-eye test) of 89.60% and 98.62% using two parameters, shape signature and its power spectrum respectively, when comparing with all the existing methods.

Eigen and Fisher Barycenter Contour for 2D Shape Classification

To achieve a good performance for shape classification, it requires both shape representation and classifier. In this paper, the so-called Eigen Barycenter Contour (EBcC) and Fisher Barycenter Contour (FBcC) techniques are presented for 2D shape classification. The representation utilizes the area of triangles at different scale level of Barycenter Contour (BcC). However, it is not invariant to starting point selection, so the phase normalization is applied. After that, we linearly project the shape feature in 3D format onto a subspace based on EBcC technique into low dimensional subspace. The FBcC, another similar method, also produces well separated classes in low dimensional subspace. Finally, the normalized cross correlation is used to measure the similarity among shapes. The experimental results demonstrate that the FBcC method outperforms the EBcC method and achieves high retrieval efficiency over other recent methods in the literature for tests on three different databases, the affine shape database, the MPEG-7 database CE-1 part B and the Kimias database.