Koichi Niijima

FPGA-based lifting wavelet processor for real-time signal detection

This paper presents the realization of a lifting wavelet processor for signal detection on a field programmable gate array (FPGA) device. This processor implements an algorithm for detecting target portions from a signal using an integer type Haar lifting wavelet transform (IHLWT), which we proposed. Since our detection algorithm is very simple because of short filter length in the IHLWT, the VLSI can be designed using a small amount of circuitry, consisting of only 6 multipliers and 9 adders. Therefore, it can realize the high-speed detection of target signals by constructing a pipeline architecture. The VLSI is designed using hardware description language (HDL) and is simulated on the FPGA in practice. The completed prototype is tested through software-generated signals and utility-sampled signals, in which test scenarios covering several kinds of electrocardiogram (ECG) signals are examined thoroughly. From the results, it is confirmed that the proposed processor can execute target signal detection from the measured ECG signals in real time.

Automatic Detection of Geomagnetic Sudden Commencement Using Lifting Wavelet Filters

This paper proposes a method for detecting geomagnetic sudden commencement (SC) from a geomagnetic horizontal (H) component by using lifting wavelet filters. Lifting wavelet filters are biorthogonal wavelet filers containing free parameters. Our method is to learn such free parameters based on some training signals which contain the SC. The learnt wavelet filters have the feature of training signals. Applying such wavelet filters to the test signals, we can detect the time when SC phenomena occurred.

Time-tunnel: visual analysis tool for time-series numerical data and its extension toward parallel coordinates

Time-tunnel proposed by Akaishi et al. is a multidimensional data analysis tool. Time-tunnel visualizes any number of time series numerical data records as individual charts each of which is displayed on an individual rectangular plane called data-wing in a 3D virtual space. Through direct manipulations on a computer screen, the user easily puts more than two data-wings overlapped together to compare their charts in order to recognize the similarity or the difference among those data records. Simultaneously a radar chart among those data at any time point is displayed to recognize the similarity and the correlation among them. Since only one chart is displayed on one data-wing, so if there is a number of data records, the user has to prepare accordingly a huge number of data-wings and practically it becomes impossible to manipulate them. To deal with this problem, the authors enhanced the functionality of time-tunnel to enable it to display more than two charts on each data-wing. In other word, the authors made each data-wing have the functionality of the parallel coordinates. This paper proposes the usefulness of this enhancement of time-tunnel by showing data analysis examples.

Treecube: visualization tool for browsing 3D multimedia data

We propose a new 3D visualization tool for hierarchical information. A 2D visualization tool for hierarchical information called treemap has already been proposed by Ben Shneiderman, et al. in 1992. In general, hierarchical information is represented as a tree structure. Treemap hierarchically lays out each node as a bounding box whose size is the same as the specific weight or attribute value of the node. After the original treemap algorithm called slice-and-dice, some extensions to it have been proposed: squarified treemap (M. Buls et al., 2000), ordered treemap (B. Shneiderman et al., 2001) and strip treemap (B.B. Bederson et al., 2002). Furthermore, quantum treemap (B.B. Bederson et al., 2002) is a quantization version of these extensions. We propose a new 3D visualization tool for hierarchical information called treecube that can be taken as a 3D extension of treemap. Especially, we show its usefulness for browsing 3D multimedia data, i.e., 2D images, 3D shape models, motion data, etc., originally stored in a file system.

Fast face detection by lifting dyadic wavelet filters

This paper presents a fast algorithm for detecting facial parts such as nose, eyes and lips in an image by using lifting dyadic wavelet filters. Free parameters in the lifting filters are learned so as to maximize the cosine of an angle between a vector whose components are the lifting filters and a vector of pixels in the facial part. Applying the learned filter to a test image, facial parts in the image is detected. In simulation, we show that our algorithm is fast and robust one for detecting facial parts from an image.

Fast, Simple, Feature Preserving and Memory Efficient Simplification of Triangle Meshes

We propose a new iterative algorithm for the automatic geometric simplification of a triangle mesh based on edge collapse transformations. The way how geometric deviation resulted from an edge collapse transformation is measured plays a basic role in inducing a priority ordering on the set of edge collapse transformations to control the simplification process. We introduce a new technique to measure this deviation based only on current simplified mesh and the new concept of accumulating the cost of collapse, which are easy to implement and involve simple computations making the algorithm computationally efficient. No geometric history is needed to be retained, so the resulting algorithm is also memory efficient. Results and numerical comparisons show that our algorithm generates simplified meshes of good visual fidelity, which compares well with those by other methods in terms of maximum and mean geometric error and it preserves the visually important features of the original mesh.

Subimage extraction by integer-type lifting wavelet transforms

This paper proposes a method for extracting subimages from a huge reference image by using lifting wavelet transforms that map integers to integers. Our integer-type lifting wavelet transform contains controllable free parameters, which are constructed based on an integer version of the Haar transform. Our learning method is to determine such free parameters using some subimages so as to remove their high frequency components in the y- and x-directions. The learnt wavelet transform has the feature of the subimages. We apply such a wavelet transform to high frequency components of a reference image and check whether they are removed or not, to detect a target subimage.

Robust lifting wavelet transform for subimage extraction

We present a novel method to extract subimages from a huge reference image by using integer-type lifting wavelet transforms. Our integer to integer lifting wavelet transform contains controllable free parameters in the lifting term, which is constructed based on an integer version of Haar wavelet transform. Such free parameters are trained following a vanishing criterion for low frequency components of query images. The trained parameters have characteristics of the query images. We apply a lifting wavelet transform with such parameters to a reference image and check whether they satisfy our vanishing criterion or not, to extract target subimages.

Similarity measure based on OBBTree for 3D model search

The demand of 3D models has been growing in the CG production and the video game industry. Many 3D models have already been created and stored. We need a 3D model database system that allows us to retrieve required 3D models quickly and accurately. In the design of 3D model database systems, there is the question which feature of 3D models is suitable as the similarity measure for 3D model searches. In this paper, the authors propose the use of OBBTree for the similarity measure and describe its usefulness by showing its experimental results and comparison results with the major method called D2. The proposed method showed better results than D2.

Fast object tracking by lifting wavelet filters

This work presents a fast object-tracking algorithm using the lifting wavelet filters. The algorithm solves a linear system of simultaneous equations of free parameters in the lifting filter and searches a small region for objects. In simulation, it is shown that the algorithm can track the object in a sequence of images.