Search

In the visual inspection, the quality assurance is difficult, because the dispersion occurs in the result by skill and fatigue degree of the inspector. Recently, a visual inspection method by image processing using deep learning has been proposed. When using deep learning, the dataset to be used is important. In this paper, we describe a method for detecting painting defects using image processing, automatically generating data for deep learning, and using these data for classification using deep learning.

Circulating tumor cells (CTCs) can be used for the early detection of cancer and determination of the effectiveness of anticancer drug treatment. Sometimes CTC grows on a culture medium which has a honeycomb texture, and existing detection methods cannot achieve satisfactory results because the honeycomb textures are more salient than the CTCs edges. Thus, a CTCs extraction method for honeycomb textures from microscope images is needed.In this paper, we propose a new algorithm to remove honeycomb texture from microscope images. We confirmed that our proposed algorithm can sufficiently remove the honeycomb texture even the periodicity of the patterns are weak. Also our method achieves a high segmentation accuracy for images where the background is more salient than the CTCs edges.

Afternoon sleepiness is a routine physiological phenomenon and it reduces arousal level, performance, and so on. It has been cleared and is often said that taking a short nap is effective to cancel the sleepiness. Sleep stage 2 is one of important factors on sleeping especially in short naps. Also, sleep spindles are important hallmarks of EEG in sleep stage 2. Therefore, it is necessary to find a spindle for analysis in sleep stage 2. In this paper, we propose a sleep spindle detector in time domain. In the proposed method, the sleep spindle waves are detected using the Fujimori's method and cumulative sum chart with occurrence tendency of candidates. Through example, the proposed method can detect the sleep spindles with 91.22 % sensitivity compared with a medical specialist.

Classifying leukocyte and examining their proportions is a very important for disease examination and diagnosis. This examination needs the knowledge of experts and a lot of time. Therefore, many automatic leukocyte images classification algorithms have been proposed. There is a method to classify 13 types blood cells using One versus One Support Vector Machine in one of them. In the conventional classify 13 types blood cells using One versus One Support Vector Machine in one of them. In the conventional method, leukocyte images are classified with the 26-dimensional feature vectors. The feature vectors are composed of 18-dimensional feature vectors relating to the color of cell and 8-dimensional feature vectors relating to the form of cell. The classification accuracy, however, is poor with these feature vectors regarding immature blood cells in this method. In this paper, we propose new feature vectors to improve the classification accuracy of immature blood cells with low classification accuracy among the leukocyte fractions. That is, we add two feature vectors in the proposed method.

Objective. The aim of this study was to evaluate the usefulness of reconstructed computed tomography (CT) images using OsiriX software in detecting wooden and bamboo foreign bodies. Methods. Four sizes of wet and dry wooden and bamboo foreign bodies were selected to be analyzed. Those in the air and in the head of edible swine were scanned with a multidetector row CT scanner. The images were evaluated with OsiriX software in the bone and the abdomen window setting as unprocessed images. Three-dimensional rendered images assigned colors and opacity by a 16-bit color look-up table (CLUT) editor in OsiriX software were evaluated as processed images. Results. In the unprocessed images, dry and wet foreign bodies in the air were not detected except a part of wet wooden foreign bodies, and all the dry and wet foreign bodies in the swine’s head mimicked air with linear shapes. In the processed images, all the dry and wet foreign bodies in the air were detected clearly, and all the wooden and some of the bamboo foreign bodies in the swine’s head were detected clearly. Conclusions. CT images processed using OsiriX software, especially with a CLUT editor, were useful in detecting wooden and bamboo foreign bodies.

The low-pass maximally flat FIR digital differentiator (low-pass MFDD) provides high accurate differentiation around the frequency ω = 0, while it reduces noises in the high frequency. Recently, authors proposed the low delay low-pass MFDD, but it can not get the steep cutoff characteristic. This paper proposes a design method of the low delay low-pass FIR digital differentiators having the flat passband and the equiripple stopband. The proposed transfer function is consist of the passband function and the stopband function. The passband function realizes the high accurate differentiation and the desired delay at ω = 0. The stopband function provides arbitrary roots in the stopband. Those roots are updated to be the optimal ones by using Remez algorithm iteratively.

In the field of digital signal processing, there are cases in which the frequency characteristics must be changed instantaneously in response to the processing objective. A filter whose characteristic can be varied instantaneously is called a variable filter. Many reports are available on the design of FIR digital filters with a variable passband. However, it appears that there are no publications on the design of an FIR digital filter in which the portion of the stopband with high attenuation can be varied. In this paper, a design method is proposed for an FIR digital filter using spectral parameters that make the portion of the stopband with high attenuation variable. For the design, the successive projection method with a simple algorithm is applied. A filter with an almost equal-ripple profile for the entire amplitude characteristic is designed by varying the spectral parameters. As a result, it is possible to obtain amplitude characteristics close to those of the filter solved by the Remez algorithm for specifications that change instantaneously from time to time. Finally, a design example is presented to confirm the effectiveness of the present approach. © 2005 Wiley Periodicals, Inc. Electron Comm Jpn Pt 3, 89(4): 34–41, 2006; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjc.20176

In this paper, we describe about the development of electric circuit exercise system that has interactiveness and effect of animation by using GUI operation, to make "Movement" of technological model easily underdtood for the students. Althogh we proposed iCass (Interactive Circuits and System Seminar) as a research similar to this paper, we advance the development of the educational support teaching material enhanced more by adding "Problem exercise" and "Learning assessment" insufficient in it.

We investigated whether images of stationary objects obtained by segmental acquisition with positron emission tomography using 2-deoxy-2-[(18)F]-fluoro-D: -glucose (FDG-PET) are of a quality equivalent to those obtained by conventional continuous acquisition.Phantoms filled with FDG and mid-abdominal regions of 18 patients who underwent FDG-PET tests were imaged by both continuous and segmental acquisition methods. The total acquisition time was set to 3 min; in the segmental acquisition mode, imaging for 15 s was repeated 12 times. Segmental images (SIs) obtained by superimposition of the reconstructed images were compared quantitatively and visually with continuous images (CIs).In all the phantom and clinical studies, SIs were never worse than CIs. The variances of the background counts of SIs were 9.8% and 13.0% less those of CIs in phantom and clinical studies, respectively. Visual assessments showed that SIs provided better detection of hot areas and superior image quality when compared to CIs.For stationary objects, the quality of images obtained by the segmental method is equivalent to that of images obtained conventionally by continuous acquisition. Moreover, under some conditions SIs provide better results than CIs.

Sampling rate conversion using the filter banks is proposed in the field of digital signal processing. However, in sampling rate conversion with a rational factor, the computational complexity may become very large. Moreover whenever the sampling rate is changed, it is necessary to design the filter again. In this paper, hence, we present a kernel with block structure for sampling rate conversion. As the filter proposed has the impulse response approximated by polynomials, it is unnecessary to redesign whenever the sampling rate is changed. The kernel proposed has the block structure that the impulse response of the sampling section since the third is represented by the polynomial used for the second sampling section. Therefore, the filter has less memory. Moreover, the filter has the advantage that it is possible to correspond to arbitrary fractional sampling rate conversion.

Several applications in digital signal processing demand variable digital filters with arbitrary variable frequency response. However, the design method of filters having variable piecewise large attenuation characteristics in the stopband is not proposed. This paper describes the design method of variable FIR digital filters having changed piecewise large attenuation characteristics in the stopband. The basic idea is to represent the coefficients of variable transfer function as the polynomials using spectral parameter. We design each filter with difference stopband attenuation by successive projections method. Then we approximate the obtained filter coefficient by the polynomials using spectral parameter in the least-square sense.

The cascade of a prefilter and an equalizer is effective for realizing an FIR digital filter with few multipliers. Generally, the number of multipliers in the entire filter decreases as the prefilter response becomes closer to a given specification. This paper proposes a new structure of pre-filters which can produce a response closer to the specification than a conventional structure. Furthermore, a new method of cascading the prefilters, and which improves both the passband and stopband characteristics, is described. In this method, a pre-filter cascaded with the preceding ones is designed so as to compensate their approximate characteristics. the small passband ripple and large stopband attenuation can be realized by using the proposed prefilters and an equalizer with only few multipliers.