• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.2
• /
• pp.219-224
• /
• 2020

Binary classification has been broadly investigated in machine learning. In addition, binary classification can be easily extended to multi class problems. To successfully utilize machine learning methods for classification tasks, preprocessing and feature extraction steps are essential. These are important steps to improve their classification performances. In this paper, we propose a new learning method based on weighted least squares. In the weighted least squares, designing weights has a significant role. Due to this necessity, we also propose a new technique to obtain weights that can achieve feature transformation. Based on this weighting technique, we also propose a method to combine the learning and feature extraction processes together to perform both processes simultaneously in one step. The proposed method shows the promising performance on five UCI machine learning data sets.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.31B no.11
• /
• pp.97-108
• /
• 1994

A modified Karhunen-Loeve transformation technique using normalization and simultaneous diagonalization of two sample covariance matrices is proposed to separate the object from the background. The transformation technique for the separation of local data structure through maximizing the ratio of sample variances between two classes was identified as a promising one for a preprocessing of multi-variate signal processing algorithms using neighborhood operators including morphological filtering. To relate the separation quality of the proposed technique to a morphological measure, average height was defined by using morphological pattern spectrum. A practical implementation of the transformation technique was tested experimentally and the theoretical results were confirmed.

• Proceedings of the IEEK Conference
• /
• 2002.07a
• /
• pp.90-93
• /
• 2002

In this work, an efficient lifting implementation of invertible deinterlacing is proposed. The invertible deinterlacing is a technique developed for intra-frame-based video coding as a preprocessing. Unlike the conventional deinterlacing, it preserves the sampling density and has the invertibility. For a special selection of filters, it is shown that the deinterlacing can be implemented efficiently by an in-place computation. It is also shown that the deinterlacing can be combined with the lifting discrete wavelet transform (BWT) employed in JPEG2000. A bit modification of the original lifting DWT is shown to provide the simultaneous implementation of deinterlacing. This fact makes the proposed technique attractive for the application to Motion-JPEG2000. The inverse transform and the reversible lifting implementation are also discussed.

• Journal of Institute of Control, Robotics and Systems
• /
• v.9 no.3
• /
• pp.227-235
• /
• 2003

We propose an algorithm for the automatic measurement of concentricity and roundness using image processing technique. The proposed measuring method consists of the preprocessing process and the measuring process. In the measuring process, two types of concentricity measurement algorithm and one type of roundness measurement algorithm are proposed. We could measure the concentricity and roundness using input image from CCD camera, without using special measurement equipment. From the experimental results, we could find that the required measurement accuracy specification is sufficiently satisfied using our proposed method.

• Proceedings of the IEEK Conference
• /
• 2006.06a
• /
• pp.697-698
• /
• 2006

In this paper, we propose a new scheme to generate multi-view images using a depth-image-based rendering (DIBR) technique. In order to improve the quality of multi-view images at newly exposed areas during mesh-based rendering, we preprocess the depth map using a Gaussian smoothing filter. Previous algorithms apply a smoothing filter to the whole depth map even if the depth map is collapsed. After extracting objects from the depth map, we apply the smoothing filter to their boundaries. Finally, we cannot only maintain the depth quality, but also generate high quality multi-view images. Experimental results show that our proposed algorithm outperforms previous works and supports an efficient depth keying technique.

• Nuclear Medicine and Molecular Imaging
• /
• v.41 no.2
• /
• pp.71-77
• /
• 2007

Advance of neuroimaging technique has greatly influenced recent brain research field. Among various neuroimaging modalities, positron emission tomography has played a key role in molecular neuroimaging though functional MRI has taken over its role in the cognitive neuroscience. As the analysis technique for PET data is more sophisticated, the complexity of the method is more increasing. Despite the wide usage of the neuroimaging techniques, the assumption and limitation of procedures have not often been dealt with for the clinician and researchers, which might be critical for reliability and interpretation of the results. In the current paper, steps of voxel-based statistical analysis of PET including preprocessing, intensity normalization, spatial normalization, and partial volume correction will be revisited in terms of the principles and limitations. Additionally, new image analysis techniques such as surface-based PET analysis, correlational analysis and multimodal imaging by combining PET and DTI, PET and TMS or EEG will also be discussed.

• Proceedings of the KIEE Conference
• /
• 2003.07d
• /
• pp.2801-2803
• /
• 2003

This paper describes a technique for human sensibility evaluation using personality-group templates of EEG(electroencephalogram). 10-channel EEGs of 5 extroverts and 5 introverts are collected in comfortable seat, uncomfortable seat and relaxed state. After preprocessing of EEG, the linear predictor coefficients are extracted and used as feature parameters. A neural network based sensibility classifier is designed and the output of the neural network is assumed as the sensibility index. Multiple templates of two personality-groups are stored and the most similar template can be selected by the proposed method. The proposed method showed the better performance than our previous results which have used ungrouped templates.

• Korean Journal of Remote Sensing
• /
• v.20 no.3
• /
• pp.207-214
• /
• 2004

This study focused on providing a methodology to utilize temporal information obtained from remotely sensed data for monitoring a wide variety of targets on the earth's surface. Generally, a methodology in understanding of global changes is composed of mapping, quantifying, and monitoring changes in the physical characteristics of land cover. The selected processing and analysis technique affects the quality of the obtained information. In this research, feature extraction methodology is proposed based on segmentation. It requires a series of processing of multitempotal images: preprocessing of geometric and radiometric correction, image subtraction/thresholding technique, and segmentation/thresholding. It results in the mapping of the change-detected areas. Here, the appropriate methods are studied for each step and especially, in segmentation process, a method to delineate the exact boundaries of features is investigated in multiresolution framework to reduce computational complexity for multitemporal images of large size.

• Journal of IKEEE
• /
• v.27 no.1
• /
• pp.59-64
• /
• 2023

Deep learning can solve various computer vision problems, but it requires a large dataset. Data augmentation technique based on image binarization for constructing large-scale datasets is proposed in this paper. By extracting features using image binarization and randomly placing the remaining pixels, new images are generated. The generated images showed similar quality to the original images and demonstrated excellent performance in deep learning models.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2021.01a
• /
• pp.11-12
• /
• 2021

소비자가 남긴 댓글이나 상품평은 상품기획의 주요 정보가 될 수 있다. 본 논문에서는 버티컬 무소음 마우스 7,300개에 대한 온라인 댓글을 딥러닝 기술인 Word2Vec을 이용하여 유사도 분석을 시행하였다. 유사도 분석결과 클릭 키워드에 대한 장점으로 소리(.975), 버튼(.972), 무게(.971)가 분석되었으며 단점은 가볍다(.959)이었다. 이는 구매 상품에 대한 소비자의 의견, 태도, 성향 및 서비스에 대한 포괄적인 의견들을 데이터화 하여 상품의 특징을 분석할 수 있는 의미있는 과정 이라고 볼 수 있다. 상품기획 프로세스에 딥러닝 기술을 통한 소비자의 감성분석자료 포함시키는 전략을 적용한다면 상품기획의 시간과 비용투자의 경제성을 높일 수 있고 나아가 빠르게 변화하는 소비자의 요구사항을 적기에 반영할 수 있을 것으로 생각된다.