• Geophysics and Geophysical Exploration
• /
• v.25 no.3
• /
• pp.140-161
• /
• 2022

Seismic data with missing traces are often obtained regularly or irregularly due to environmental and economic constraints in their acquisition. Accordingly, seismic data interpolation is an essential step in seismic data processing. Recently, research activity on machine learning-based seismic data interpolation has been flourishing. In particular, convolutional neural network (CNN) and generative adversarial network (GAN), which are widely used algorithms for super-resolution problem solving in the image processing field, are also used for seismic data interpolation. In this study, CNN-based algorithm, U-Net and GAN-based algorithm, and conditional Wasserstein GAN (cWGAN) were used as seismic data interpolation methods. The results and performances of the methods were evaluated thoroughly to find an optimal interpolation method, which reconstructs with high accuracy missing seismic data. The work process for model training and performance evaluation was divided into two cases (i.e., Cases I and II). In Case I, we trained the model using only the regularly sampled data with 50% missing traces. We evaluated the model performance by applying the trained model to a total of six different test datasets, which consisted of a combination of regular, irregular, and sampling ratios. In Case II, six different models were generated using the training datasets sampled in the same way as the six test datasets. The models were applied to the same test datasets used in Case I to compare the results. We found that cWGAN showed better prediction performance than U-Net with higher PSNR and SSIM. However, cWGAN generated additional noise to the prediction results; thus, an ensemble technique was performed to remove the noise and improve the accuracy. The cWGAN ensemble model removed successfully the noise and showed improved PSNR and SSIM compared with existing individual models.

• The Journal of the Acoustical Society of Korea
• /
• v.36 no.5
• /
• pp.338-344
• /
• 2017

In this paper, an interpolation method of HRTF (Head-Related Transfer Function) is proposed for small-sized measurement data set, especially. The proposed algorithm is based on acoustic modeling of HRTFs, and the algorithm tries to interpolate the HRTFs via estimation the model coefficients. However, the estimation of the model coefficients is hard if there is lack of measurement points, so the algorithm solves the problem by a data augmentation using the VBAP (Vector Based Amplitude Panning). Therefore, the proposed algorithm consists of two steps, which are data augmentation step based on VBAP and model coefficients estimation step by least squares method. The proposed algorithm was evaluated by a simulation with a measured data from CIPIC (Center for Image Processing and Integrated Computing) HRTF database, and the simulation results show that the proposed algorithm reduces mean-squared error by 1.5 dB ~ 4 dB than the conventional algorithms.

• 한국전산유체공학회:학술대회논문집
• /
• 2005.10a
• /
• pp.131-134
• /
• 2005

Conventional high order interpolation schemes are limitative in several aspects mainly because they need data of neighboring cells at the reconstruction step. However, discontinuous Galerkin method and spectral volume method, two high order flux schemes which will be analyzed and compared in this paper, have an important benefit that they are not necessary to determine the flow gradients from data of neighboring cells or elements. These two schemes construct polynomial of variables within a cell so that even near wall or discontinuity, the high order does not deteriorate.

• Journal of Broadcast Engineering
• /
• v.15 no.2
• /
• pp.265-279
• /
• 2010

A color image requires at least three color channels to obtain the full color image. However the image sensor obtains only the intensity of the brightness, that is, three image sensors are required for every pixel to capture the full color image. Since the image sensor is quiet expensive, most of digital still cameras adopt single image sensor array with color filter array (CFA) to reduce the size and the cost. Since the image obtained using single sensor array has only one color component per pixel, we need to reconstruct the missing two color components to obtain the full color image. We call this process as color filter interpolation or demosaicking. In this paper, demosaicking algorithm composed of two large step is proposed. Proposed algorithm is combined with several different algorithms such as Edge-directed demosaicking, Second-order gradients as correction terms, Smooth hue transition Interpolation, etc. The simulation results show that the proposed algorithm performs much better than the state-of-the-art demosaicking algorithms in terms of both subjective and objective qualities.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.47 no.5
• /
• pp.87-93
• /
• 2010

This paper presents a high resolution, wide input range 2-step time-to-digital converter used in digital PLL. TDC is used to compare the DPLL output frequency with reference frequency and should be implemented with high resolution to improve the phase noise of DPLL. The conventional TDC consists of delay line realized inverters, whose resolution is determined by delay time of inverter and transistor size, resulting in limited resolution. In this paper, 2-step TDC with phase-interpolation and Time Amplifier is proposed to meet the high resolution and wide input range by implement the delay time less than an inverter delay. The gain of Time Amplifier is improved by using the delay time difference between two inverters. It is implemented in $0.13{\mu}m$ CMOS process and the die area is $800{\mu}m{\times}850{\mu}m$ Current consumption is 12 mA at the supply voltage of 1.2 V. The resolution and input range of the proposed TDC are 0.357 ps and 200 ps, respectively.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.4 no.2
• /
• pp.407-416
• /
• 2000

Adaptive digital filtering and noise cancelling technique using a tree structured filter bank are presented to reduce a undesirable aliasing due to the decimation of filtered output and improve the performance in terms of mean-square error and the convergence speed using a aliasing canceller. A signal is split into two subband by analysis filter bank and decimated by decimator and reconstructed by interpolation technique and synthesis filter bank. A variable step-size LMS algorithm is used to improve the convergence speed in case of existing the measurement noise in desired input of filter. It is shown by computer simulation that the proposed subband structure in this paper is superior to conventional subband filter structure in terms of mean-square error and convergence speed.

• Journal of the Korean Society for Precision Engineering
• /
• v.11 no.6
• /
• pp.117-126
• /
• 1994

This paper describes an application step of a $R^{-{\theta}}$ method which measures circular movements in machining centers. The detection of composite errors of circular movements and a high precision cutting method in machining centers were investigated by the analysis of data measured by $R^{\theta }$method which can detect the rotating angle and is applicable to variable measuring radius. When the error by squareness error and unbalance of position-loop-gain were mixed, the detection method of each error was proposed. Although the errors by squarenss error and backlash compensation were mixed, the errors by squareness error be detected. If the errors by unbalance of position-loop-gain and backlash compensation were mixed, the errors by unbalance of position-loop-gain could not detected. A high precision cutting mehod, which uses the NC program compensated by using feed-back data from error measured by the $R^{\theta }$method, was proposed.

• Proceedings of the KSRS Conference
• /
• 2003.11a
• /
• pp.1137-1139
• /
• 2003

This paper proposed a practical method for building detection and extraction using airborne laser scanning data. The proposed method consists mainly of two processes: low and high level processes. The major distinction from the previous approaches is that we introduce a concept of pseudogrid (or binning) into raw laser scanning data to avoid the loss of information and accuracy due to interpolation as well as to define the adjacency of neighboring laser point data and to speed up the processing time. The approach begins with pseudo-grid generation, noise removal, segmentation, grouping for building detection, linearization and simplification of building boundary , and building extraction in 3D vector format. To achieve the efficient processing, each step changes the domain of input data such as point and pseudo-grid accordingly. The experimental results shows that the proposed method is promising.

• Journal of Korea Multimedia Society
• /
• v.21 no.12
• /
• pp.1467-1472
• /
• 2018

For the problem that the face image of surveillance video cannot be accurately identified due to the low resolution, this paper proposes a low resolution face recognition solution based on convolutional neural network model. Convolutional Neural Networks (CNN) model for multi-scale input The CNN model for multi-scale input is an improvement over the existing "two-step method" in which low-resolution images are up-sampled using a simple bi-cubic interpolation method. Then, the up sampled image and the high-resolution image are mixed as a model training sample. The CNN model learns the common feature space of the high- and low-resolution images, and then measures the feature similarity through the cosine distance. Finally, the recognition result is given. The experiments on the CMU PIE and Extended Yale B datasets show that the accuracy of the model is better than other comparison methods. Compared with the CMDA_BGE algorithm with the highest recognition rate, the accuracy rate is 2.5%~9.9%.

• Journal of Ship and Ocean Technology
• /
• v.10 no.2
• /
• pp.24-36
• /
• 2006

Ship hullform is usually designed with a curve network, and smooth hullform surfaces are supposed to be generated by filling in (or interpolating) the curve network with appropriate surface patches. Tensor-product surfaces such as B-spline and $B\'{e}zier$ patches are typical representations to this interpolating problem. However, they have difficulties in representing the surfaces of irregular topological type which are frequently appeared in the fore- and after-body of ship hullform curve network. In this paper, we proposed a method that can automatically generate discrete $G^1$ continuous B-spline surfaces interpolating given curve network of ship hullform. This method consists of three steps. In the first step, given curve network is reorganized to be of two types: boundary curves and reference curves of surface patches. Especially, the boundary curves are specified for their surface patches to be rectangular or triangular topological type that can be represented with tensor-product (or degenerate) B-spline surface patches. In the second step, surface fitting points and cross boundary derivatives are estimated by constructing virtual iso-parametric curves at discrete parameters. In the last step, discrete $G^1$ continuous B-spline surfaces are generated by surface fitting algorithm. Finally, several examples of resulting smooth hullform surfaces generated from the curve network data of actual ship hullform are included to demonstrate the quality of the proposed method.