• Journal of Korea Society of Digital Industry and Information Management
• /
• v.16 no.1
• /
• pp.55-65
• /
• 2020

With the development of display hardware, image interpolation techniques have been used in various fields such as image zooming and medical imaging. Traditional image interpolation methods, such as bi-linear interpolation, bi-cubic interpolation and edge direction-based interpolation, perform interpolation in the spatial domain. Recently, interpolation techniques in the discrete cosine transform or wavelet domain are also proposed. Using these various existing interpolation methods and machine learning, we propose decision tree classification-based image interpolation methods. In other words, this paper is about the method of adaptively applying various existing interpolation methods, not the interpolation method itself. To obtain the decision model, we used Weka's J48 library with the C4.5 decision tree algorithm. The proposed method first constructs attribute set and select classes that means interpolation methods for classification model. And after training, interpolation is performed using different interpolation methods according to attributes characteristics. Simulation results show that the proposed method yields reasonable performance.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.16 no.3
• /
• pp.49-58
• /
• 2020

Image interpolation, a technology that converts low resolution images into high resolution images, has been widely used in various image processing fields such as CCTV, web-cam, and medical imaging. This technique is based on the fact that the statistical distributions of the white Gaussian noise and the difference between the interpolated image and the original image is similar to each other. The proposed algorithm is composed of three steps. In first, the interpolated image is derived by random image interpolation. In second, we derive weighting functions that are used to apply non-local mean filtering. In the final step, the prediction error is corrected by performing non-local mean filtering by applying the selected weighting function. It can be considered as a post-processing algorithm to further reduce the prediction error after applying an arbitrary image interpolation algorithm. Simulation results show that the proposed method yields reasonable performance.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37 no.6A
• /
• pp.464-472
• /
• 2012

In this paper, we propose a new super-resolution algorithm that uses successive frames by applying the block matching motion estimation algorithm. Usually, single frame super-resolution algorithms are based on probability or discrete wavelet transform (DWT) approach to extract high-frequency components of the input image, but only limited information is available for these algorithms. To solve this problem, various multiple-frame based super-resolution algorithms are proposed. The accuracy of registration between frames is a very important factor for the good performance of an algorithm. We therefore propose an algorithm using 6-Tap FIR filter to increase the accuracy of the image registration with sub-pixel unit. Proposed algorithm shows better performance than other conventional interpolation based algorithms such as nearest neighborhood, bi-linear and bi-cubic methods and results in about the same image quality as DWT based super-resolution algorithm.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2022.06a
• /
• pp.1226-1228
• /
• 2022

FHD 이상을 넘어선 UHD급의 고해상도 동영상 콘텐츠의 수요 및 공급이 증가함에 따라 전반적인 산업 영역에서 네트워크 자원을 효율적으로 이용하여 동영상 콘텐츠를 제공하는 데에 관심을 두게 되었다. 기존 방법을 통한 bi-cubic, bi-linear interpolation 등의 방법은 딥 러닝 기반의 모델에 비교적 인풋 이미지의 특징을 잘 잡아내지 못하는 결과를 나타내었다. 딥 러닝 기반의 초 해상화 기술의 경우 기존 방법과 비교 시 연산을 위해 더 많은 자원을 필요로 하므로, 이러한 사용 조건에 따라 본 논문은 초 해상화가 가능한 딥 러닝 모델을 경량화 기법을 사용하여 기존에 사용된 모델보다 비교적 적은 자원을 효율적으로 사용할 수 있도록 연구 개발하는 데 목적을 두었다. 연구방법으로는 structure pruning을 이용하여 모델 자체의 구조를 경량화 하였고, 학습을 진행해야 하는 파라미터를 줄여 하드웨어 자원을 줄이는 연구를 진행했다. 또한, Residual Network의 개수를 줄여가며 PSNR, LPIPS, tOF등의 결과를 비교했다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2016.06a
• /
• pp.189-190
• /
• 2016

일반 영상의 영상확대를 위한 다양한 알고리즘이 존재한다. 하지만 적외선 열화상 영상의 경우 일반영상과 다른 특성을 가지고 있기 때문에 적외선 영상을 위한 영상 확대 알고리즘이 필요하다. 따라서 적외선 영상이 일반영상에 비해 디테일이 없다는 특성을 고려하여 복잡한 알고리즘을 적용시키기 보다는 ADRC 와 같은 단순한 분류 기법을 활용하여 LR-HR 패치를 분류하고 학습된 데이터를 이용하여 영상확대 알고리즘에 적용하였다. 알고리즘의 성능 향상을 위해 학습과정에 전처리 과정을 추가하여 합성과정에서 추가적인 연산량의 증가 없이 확대 영상의 선명도를 향상시키고자 하였다. 또한 확대된 적외선 영상이 동일 해상도의 가시광영상에 비해 선명도가 떨어진다는 점을 고려하여 확대된 적외선 영상에 가시광영상의 고주파 정보를 합성시켜 이전보다 영상의 선명도를 더 향상시키고자 하였다. 이와 같은 방법으로 영상 확대 알고리즘만 수행하였을 때 통상적인 영상확대 기법인 bi-cubic interpolation 기법보다 JNB 수치가 평균 0.0727 만큼 높은 결과를 확인할 수 있었고 가시광영상과 융합하였을 때 이전보다 평균 0.0742 만큼 더 선명해진 영상을 얻었다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2015.11a
• /
• pp.152-153
• /
• 2015

일반 영상의 영상확대를 위한 다양한 알고리즘이 존재한다. 하지만 IR 영상의 경우 일반영상과 다른 특성을 가지고 있기 때문에 IR 영상을 위한 영상 확대 알고리즘이 필요하다. 따라서 IR 영상이 일반영상에 비해 디테일이 없다는 특성을 고려하여 복잡한 알고리즘을 적용시키기 보다는 ADRC[1]와 같은 단순한 분류 기법을 활용하여 LR-HR 패치를 분류하고 학습된 데이터를 이용하여 영상확대 알고리즘에 적용하였다. 또한 알고리즘의 성능을 향상시키기 위해 학습과정에 전처리 과정을 추가하여 알고리즘 작동 시 연산량의 증가 없이 확대 영상의 선명도를 향상시키고자 하였다. 이와 같은 방법으로 영상 확대 알고리즘을 수행하였을 때 통상적인 영상확대 기법인 bi-cubic interpolation 기법보다 CPBD 수치가 평균 0.0527 만큼 높은 결과를 확인할 수 있었고 전처리 과정을 추가하였을 때 이전보다 평균 0.0119 만큼 더 선명해진 영상을 얻었다.

• 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 Power System Engineering
• /
• v.10 no.4
• /
• pp.56-64
• /
• 2006

This paper shows the development process of a straight-type five-hole pressure probe for measuring three-dimensional flow velocity components. The data reduction method using a bi-cubic curve-fitting program in a new calibration map was introduced in this study. This new calibration map can be applied up to the application angle, ${\pm}55^{\circ}$ of a probe. As a result, for the application angle of ${\pm}45^{\circ}$, an error for yaw and pitch angles appeared from $-1.76^{\circ}\;to\;1.83^{\circ}$ and from $-1.91^{\circ}\;to\;1.75^{\circ}$, respectively. Moreover, an error for a vector magnitude and a static pressure compared with a dynamic one showed from -7.83% to 4.87% and from -0.73 to 0.77, respectively. Even though this data reduction method showed unsatisfactory errors in a vector magnitude, it resulted in an easy and simple application method. Especially, when it was applied to an actual flow field including a swirling flow, a good result came out on the whole. However, in order to obtain a better result, it is thought that a more sophisticated interpolation method needs to be introduced.

• The Bulletin of The Korean Astronomical Society
• /
• v.44 no.1
• /
• pp.52.1-52.1
• /
• 2019

The Helioseismic and Magnetic Imager (HMI) is the instrument of Solar Dynamics Observatory (SDO) to study the magnetic field and oscillation at the solar surface. The HMI image is not enough to analyze very small magnetic features on solar surface since it has a spatial resolution of one arcsec. Super resolution is a technique that enhances the resolution of a low resolution image. In this study, we use a method for enhancing the solar image resolution using a Deep-learning model which generates a high resolution HMI image from a low resolution HMI image (4 by 4 binning). Deep learning networks try to find the hidden equation between low resolution image and high resolution image from given input and the corresponding output image. In this study, we trained a model based on a very deep residual channel attention networks (RCAN) with HMI images in 2014 and test it with HMI images in 2015. We find that the model achieves high quality results in view of both visual and measures: 31.40 peak signal-to-noise ratio(PSNR), Correlation Coefficient (0.96), Root mean square error (RMSE) is 0.004. This result is much better than the conventional bi-cubic interpolation. We will apply this model to full-resolution SDO/HMI and GST magnetograms.

• The Bulletin of The Korean Astronomical Society
• /
• v.44 no.2
• /
• pp.70.4-70.4
• /
• 2019

Image super-resolution (SR) is a technique that enhances the resolution of a low resolution image. In this study, we use three SR models (RCAN, ProSRGAN and Bicubic) for enhancing solar SDO/HMI magnetograms using deep learning. Each model generates a high resolution HMI image from a low resolution HMI image (4 by 4 binning). The pixel resolution of HMI is about 0.504 arcsec. Deep learning networks try to find the hidden equation between low resolution image and high resolution image from given input and the corresponding output image. In this study, we trained three models with HMI images in 2014 and test them with HMI images in 2015. We find that the RCAN model achieves higher quality results than the other two methods in view of both visual aspects and metrics: 31.40 peak signal-to-noise ratio(PSNR), Correlation Coefficient (0.96), Root mean square error (RMSE) is 0.004. This result is also much better than the conventional bi-cubic interpolation. We apply this model to a full-resolution SDO/HMI image and compare the generated image with the corresponding Hinode NFI magnetogram. As a result, we get a very high correlation (0.92) between the generated SR magnetogram and the Hinode one.