• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.16 no.2
• /
• pp.113-122
• /
• 2016

In this paper, we suggest a approach which comprises fast Fourier transform inversion by wavelet noise attenuation. It represents an inverse filtering by adopting a factor into the Wiener filtering, and the optimal factor is chosen to minimize the overall mean squared error. in order to apply the Wiener filter, we have to compute the power spectrum of original image from the corrupted figure. Since the Wiener filtering contains the inverse filtering process, it expands the noise when the blurring filter is not invertible. To remove the large noises, the best is to remove the noise using wavelet threshold. Wavelet noise attenuation steps are consisted of inverse filtering and noise reduction by Wavelet functions. experimental results have not outperformed the other methods over the overall restoration performance.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2019.05a
• /
• pp.236-237
• /
• 2019

In this paper, we propose a wavelet - based image noise reduction algorithm. We develop wavelet transform of existing Mallat Tree method. First, we propose a multiple filtering method. Maximizes the energy concentration characteristic of the wavelet transform considering the energy of each subband in the wavelet domain. We apply the proposed multiple filtering to the noise image. Finds energy subbands that can not be seen in normal images and removes them to remove noise.

• Journal of the Korean Society of Radiology
• /
• v.13 no.3
• /
• pp.453-458
• /
• 2019

In this study we analyzed the tendency of the image characteristic by changing filtering factor for the proposed fast non local means (FNLM) noise reduction algorithm with designed Male Adult mesh (MASH) phantom through Geant4 application for tomographic emission (GATE) simulation program. To accomplish this purpose, MASH phantom for human copy was designed through the GATE simulation program. In addition, we acquired degraded image by adding Gaussian noise with a value of 0.005 using the MATALB program in MASH phantom. Moreover, in degraded image, the FNLM noise reduction algorithm was applied by changing the filtering factors, which set to 0.005, 0.01, 0.05, 0.1, 0.5, and 1.0 value, respectively. To quantitatively evaluate, the coefficient of variation (COV), signal to noise ratio (SNR), and contrast to noise ratio (CNR) were calculated in reconstructed images. Results of the COV, SNR and CNR were most improved in image with a filtering factor of 0.05 value. Especially, the COV was decreased with increasing filtering factor, and showed nearly constant values after 0.05 value of the filtering factor. In addition, SNR and CNR were showed that improvement with increasing filtering factor, and deterioration after 0.05 value of the filtering factor. In conclusion, we demonstrated the significance of setting the filtering factor when applying the FNLM noise reduction algorithm in degraded image.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2019.06a
• /
• pp.225-226
• /
• 2019

본 논문에서는 JPEG Pleno에서 표준 데이터세트로 제공되고 있는 랜덤 위상 홀로그램의 압축 방법을 제시한다. 일반적으로 영상의 노이즈 성분은 압축의 효율을 크게 떨어트린다. 스페클 노이즈가 심한 홀로그램의 경우 일반 영상에 비해 압축률 대비 화질이 좋지 않다. 홀로그램의 시각화를 위한 랜덤 위상의 추가는 스페클 노이즈와 더블어 홀로그램의 압축 효율을 더욱이 떨어트린다. 랜덤 위상 홀로그램에 웨이블릿 기반의 다중 필터링 방법을 적용한다. 다중 필터링 방법은 홀로그램의 특성을 고려하여 에너지 집중도를 최대한 높이는 방법이다. 에너지 집중도가 높을수록 압축 효율이 좋은 Zero-Tree 방식의 압축을 통해 홀로그램을 압축하고 압축률 대비 정량적 화질평가로 그 효율을 분석한다.

• Science of Emotion and Sensibility
• /
• v.25 no.1
• /
• pp.67-78
• /
• 2022

Interest in bio-signal monitoring of wearable devices is increasing significantly as the next generation needs to develop new devices to dominate the global market of the information and communication technology industry. Accordingly, this research developed a resistive textile strain sensor through a wetting process in a single-wall carbon nanotube dispersion solution using an E-Band with low hysteresis. To measure the resistance signal in the E-Band to which electrical conductivity is applied, a universal material tester, an Arduino, and LCR meters that are microcontroller units were used to measure the resistance change according to the tensile change. To effectively handle various noises generated due to the characteristics of the fabric textile strain sensor, the filter performance of the sensor was evaluated using the moving average filter, Savitsky-Golay filter, and intermediate filters of signal processing. As a result, the reliability of the filtering result of the moving average filter was at least 89.82% with a maximum of 97.87%, and moving average filtering was suitable as the noise filtering method of the textile strain sensor.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.45 no.2
• /
• pp.74-80
• /
• 2008

Spatio-temporal filters are capable of reducing noise from moving pictures, which cannot be dealt with by spatial filters. However, the algorithm is too complicated to be realized as hardware. We implemented a real-time spatio-temporal noise reduction system, using at most three frames, based upon adaptive mean filter algorithm. Some factors which interfere with hardware implementation were modified. Noise estimated from the previous frame was used to filter the current frame so that filtering could be conducted in parallel with noise estimation. This speeds up the system thereby enabling real-time execution. The form of filtering windows was also modified to facilitate synchronization. The proposed structure was implemented on Virtex 4 XC4VLX60, occupying 66% of total slices with 80MHz of the maximum operation frequency.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.7
• /
• pp.19-27
• /
• 2023

In the process of collecting social reviews, a number of noise reviews irrelevant to a given search keyword can be included in the search results. To filter out such reviews, machine learning can be used. However, if the number of reviews is insufficient for a target place to be analyzed, filtering accuracy can be degraded due to the lack of training data. To resolve this issue, we propose a supervised learning method to improve accuracy of the noise review filtering for the places with insufficient reviews. In the proposed method, training is not performed by an individual place, but by a group including several places with similar characteristics. The classifier obtained through the training can be used for the noise review filtering of an arbitrary place belonging to the group, so the problem of insufficient training data can be resolved. To verify the proposed method, a noise review filtering model was implemented using LSTM and BERT, and filtering accuracy was checked through experiments using real data collected online. The experimental results show that the accuracy of the proposed method was 92.4% on the average, and it provided 87.5% accuracy when targeting places with less than 100 reviews.

• Annual Conference on Human and Language Technology
• /
• 2013.10a
• /
• pp.64-69
• /
• 2013

스마트폰의 대중화로 인하여 앱 마켓 시장이 급속도로 성장하였다. 이로 인하여 하루에도 수십개의 새로운 앱들이 출시되고 있다. 이러한 앱 마켓 시장의 급격한 성장으로 인해 사용자들은 자신이 흥미를 가질만한 앱들을 선택하는데 큰 어려움을 겪고 있어 앱 추천 방법에 대한 연구에 많은 관심이 집중되고 있다. 기존 연구에서 협력 필터링 기반의 추천 방법들을 제안하였으나 이는 콜드 스타트 문제를 지니고 있다. 이와는 달리 컨텐츠 기반 필터링 방식은 콜드 스타트 문제를 효율적으로 해소할 수 있는 방법이지만 앱설명에는 광고, 공지사항등 실질적으로 앱의 특징과는 무관한 노이즈들이 다수 존재하고 이들은 앱 사이의 유사관계를 파악하는데 방해가 된다. 본 논문에서는 이런 문제를 해결하기 위하여 앱 설명에서 노이즈에 해당하는 설명들을 자동으로 제거할 수 있는 모델을 제안한다. 제안하는 모델은 모바일 앱 설명을 구성하고 있는 각 문단을 LDA로 학습된 토픽들의 비율로 나타내고 이들을 분류문제에서 우수한 성능을 보이는 SVM을 이용하여 분류한다. 실험 결과에 따르면 본 논문에서 제안한 방법은 기존에 문서 분류에 많이 사용되는 Bag-of-Word 표현법에 기반한 문서 표현 방식보다 더 나은 분류 성능을 보였다.

• Proceedings of the KIEE Conference
• /
• 2002.07d
• /
• pp.2777-2779
• /
• 2002

본 논문에서는 영상안에서의 노이즈를 제거하기 위한 방법과 영상을 압축하기 위한 방법을 제안하였다. 영상을 필터링하기 위한 방법으로 해상도의 손상 없이 영상의 신호대잡음비(SNR)를 개선시킬 수 있는 국부 형태 적응 필터링을 제안하였다. 제안한 알고리즘에서는 템플릿 형태가 고정되어 있는 기존의 필터링 방법 대신에 다중 템플릿들을 정의하였다. 적응 템플릿 필터링을 자기공명영상에 적용할 때 기존의 필터링 방법들에 비하여 향상된 결과를 얻을 수 있으나. $T_1$ 영상과 같이 비교적 작은 동적 범위를 가진 영상에서는 에지에서 계단모양의 artifact가 발견되곤 한다. 본 논문에서는 다중 성분을 갖는 복셀들을 선별하여 이들에 대해서는 가장 큰 크기의 템플릿을 할당함으로써 artifact를 제거하는 방법을 제안하였다. 영상 압축에 있어서는 두 가지 모델이 제안되었다. 첫 번째로, 향상된 정수 기반 웨이블렛 변환을 사용한 무손실에 가까운 압축을 제안하였으며, 두 번째로, 완전 복원이 가능한 정수 기반 웨이블렛 변환을 사용한 통합된 유/무손실 압축을 제안하였다. 모의 실험에서, 제안된 알고리즘에 의해 재구성된 영상들은 부동 소수점 기반 웨이블렛 변환과 JPEG에 의해 재구성된 영상들에 비해 높은 신호대잡음비를 보였다.

• Journal of Biomedical Engineering Research
• /
• v.19 no.2
• /
• pp.119-128
• /
• 1998

This paper describes a filtering algorithm which employs apriori information of SPECT lesion detectability potential for the filtering of degraded projection images prior to the backprojection reconstruction. In this algorithm, we determined m minimum detectable lesion sized(MDLSs) by assuming m object contrasts uniformly-chosen in the range of 0.0-1.0, based on a signal/noise model which provides the capability potential of SPECT in terms of physical factors. A best estimate of given projection image is attempted as a weighted combination of the subimages from m optimal filters whose design is focused on maximizing the local S/N ratios for the MDLS-lesions. These subimages show relatively larger resolution recovery effect and relatively smaller noise reduction effect with the decreased MDLS, and the weighting on each subimage was controlled by the difference between the subimage and the maximum-resolution-recovered projection image. The proposed filtering algoritym was tested on SPECT image reconstruction problems, and produced good results. Especially, this algorithm showed the adaptive effect that approximately averages the filter outputs in homogeneous areas and sensitively depends on each filter strength on contrast preserving/enhancing in textured lesion areas of the reconstructed image.