• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.10 no.6
• /
• pp.2527-2545
• /
• 2016

Measuring the similarity of given samples is a key problem of recognition, clustering, retrieval and related applications. A number of works, e.g. kernel method and metric learning, have been contributed to this problem. The challenge of similarity learning is to find a similarity robust to intra-class variance and simultaneously selective to inter-class characteristic. We observed that, the similarity measure can be improved if the data distribution and hidden semantic information are exploited in a more sophisticated way. In this paper, we propose a similarity learning approach for retrieval and recognition. The approach, termed as LDA-FEK, derives free energy kernel (FEK) from Latent Dirichlet Allocation (LDA). First, it trains LDA and constructs kernel using the parameters and variables of the trained model. Then, the unknown kernel parameters are learned by a discriminative learning approach. The main contributions of the proposed method are twofold: (1) the method is computationally efficient and scalable since the parameters in kernel are determined in a staged way; (2) the method exploits data distribution and semantic level hidden information by means of LDA. To evaluate the performance of LDA-FEK, we apply it for image retrieval over two data sets and for text categorization on four popular data sets. The results show the competitive performance of our method.

• Journal of Biomedical Engineering Research
• /
• v.43 no.5
• /
• pp.341-352
• /
• 2022

Medical image segmentation is the most important task in radiation therapy. Especially, when segmenting medical images, the liver is one of the most difficult organs to segment because it has various shapes and is close to other organs. Therefore, automatic segmentation of the liver in computed tomography (CT) images is a difficult task. Since tumors also have low contrast in surrounding tissues, and the shape, location, size, and number of tumors vary from patient to patient, accurate tumor segmentation takes a long time. In this study, we propose a method algorithm for automatically segmenting the liver and tumor for this purpose. As an advantage of setting the boundaries of the tumor, the liver and tumor were automatically segmented from the CT image using the 2D CoordConv DeepLab V3+ model using the CoordConv layer. For tumors, only cropped liver images were used to improve accuracy. Additionally, to increase the segmentation accuracy, augmentation, preprocess, loss function, and hyperparameter were used to find optimal values. We compared the CoordConv DeepLab v3+ model using the CoordConv layer and the DeepLab V3+ model without the CoordConv layer to determine whether they affected the segmentation accuracy. The data sets used included 131 hepatic tumor segmentation (LiTS) challenge data sets (100 train sets, 16 validation sets, and 15 test sets). Additional learned data were tested using 15 clinical data from Seoul St. Mary's Hospital. The evaluation was compared with the study results learned with a two-dimensional deep learning-based model. Dice values without the CoordConv layer achieved 0.965 ± 0.01 for liver segmentation and 0.925 ± 0.04 for tumor segmentation using the LiTS data set. Results from the clinical data set achieved 0.927 ± 0.02 for liver division and 0.903 ± 0.05 for tumor division. The dice values using the CoordConv layer achieved 0.989 ± 0.02 for liver segmentation and 0.937 ± 0.07 for tumor segmentation using the LiTS data set. Results from the clinical data set achieved 0.944 ± 0.02 for liver division and 0.916 ± 0.18 for tumor division. The use of CoordConv layers improves the segmentation accuracy. The highest of the most recently published values were 0.960 and 0.749 for liver and tumor division, respectively. However, better performance was achieved with 0.989 and 0.937 results for liver and tumor, which would have been used with the algorithm proposed in this study. The algorithm proposed in this study can play a useful role in treatment planning by improving contouring accuracy and reducing time when segmentation evaluation of liver and tumor is performed. And accurate identification of liver anatomy in medical imaging applications, such as surgical planning, as well as radiotherapy, which can leverage the findings of this study, can help clinical evaluation of the risks and benefits of liver intervention.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.28 no.6D
• /
• pp.915-922
• /
• 2008

Impervious surface is an important index for the estimation of urbanization and the assessment of environmental change. In addition, impervious surface influences on short-term rainfall-runoff model during rainy season in hydrology. Recently, the necessity of impervious surface estimation is increased because the effect of impervious surface is increased by rapid urbanization. In this study, impervious surface estimation is performed by using remote sensing image such as Landsat-7 ETM+image with $30m{\times}30m$ spatial resolution and satellite image with $1m{\times}1m$ spatial resolution based on Jungnangcheon basin. A tasseled cap transformation and NDVI(normalized difference vegetation index) transformation are applied to Landsat-7 ETM+ image to collect various predict variables. Moreover, the training data sets are collected by overlaying between Landsat-7 ETM+ image and satellite image, and CART(classification and regression tree) is applied to the training data sets. As a result, impervious surface prediction model is consisted and the impervious surface map is generated for Jungnangcheon basin.

• Nuclear Engineering and Technology
• /
• v.53 no.2
• /
• pp.522-531
• /
• 2021

We describe a new deep learning model - Convolutional eXtreme Gradient Boosting (ConvXGB) for classification problems based on convolutional neural nets and Chen et al.'s XGBoost. As well as image data, ConvXGB also supports the general classification problems, with a data preprocessing module. ConvXGB consists of several stacked convolutional layers to learn the features of the input and is able to learn features automatically, followed by XGBoost in the last layer for predicting the class labels. The ConvXGB model is simplified by reducing the number of parameters under appropriate conditions, since it is not necessary re-adjust the weight values in a back propagation cycle. Experiments on several data sets from UCL Repository, including images and general data sets, showed that our model handled the classification problems, for all the tested data sets, slightly better than CNN and XGBoost alone and was sometimes significantly better.

• Journal of Advanced Marine Engineering and Technology
• /
• v.23 no.2
• /
• pp.259-269
• /
• 1999

A new algorithm for measuring 3-D velocity components of high speed flows were developed using a digital image processing technique. The measuring system consists of three CCD cameras an optical instrument called AOM a digital image grabber and a host computer. The images of mov-ing particles arranged spatially on a rotation plate are taken by two or three CCD cameras and are recorderd onto the image grabber or a video tape recoder. The three-dimensionl velocity com-ponents of the particles are automatically obtained by the developed algorithm In order to verify the validity of this technique three-dimensional velocity data sets obtained from a computer simu-lation of a backward facing step flow were used as test data for the algorithm. an uncertainty analysis associated with the present algorithm is systematically evaluated, The present technique is proved to be used as a tookl for the measurement of unsteady three-dimensional fluid flows.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.34 no.11A
• /
• pp.870-881
• /
• 2009

The rapid development of healthcare information management for 3D medical digital library, 3D PACS and 3D medical diagnosis has addressed security issues with medical IT technology. This paper presents multiple 3D medical image data for protection, authentication, indexing and diagnosis information hiding applied to healthcare information management. The proposed scheme based on POCS watermarking embeds the robust watermark for doctor's digital signature and information retrieval indexing key to the distribution of vertex curvedness and embeds the fragile watermark for diagnosis information and authentication reference message to the distance difference of vertex. The multiple embedding process designs three convex sets for robustness, fragileness and invisibility and projects 3D medical image data onto three convex sets alternatively and iteratively. Experimental results confirmed that the proposed scheme has the robustness and fragileness to various 3D geometric and mesh modifiers at once.

• Proceedings of the KSRS Conference
• /
• 2003.11a
• /
• pp.1393-1395
• /
• 2003

In this research, we presented a data integration, which integrates ultra high resolution images and complementary data for 3D building reconstruction. In our method, as the ultra high resolution image, Three Line Sensor (TLS) images are used in combination with 2D digital maps, DSMs and both of them. Reconstructed 3D buildings, correctness rate and the accuracy of results were presented. As a result, optimized combination scheme of data sets , sensors and methods was proposed.

• Journal of the Korean BIBLIA Society for library and Information Science
• /
• v.24 no.4
• /
• pp.75-90
• /
• 2013

Given the context that users are actively using social media with multimedia embedded information, the purpose of this study is to demonstrate how images are used within Twitter messages, especially in influential and favorited messages. In order to achieve the purpose of this study, the top 200 influential and favorited messages with images were selected out of 1,589 tweets related to "Boston bombing" in April 2013. The characteristics of the message, image use, and user are analyzed and compared. Two phases of the analysis were conducted on three data sets containing the top 200 influential messages, top 200 favorited messages, and general messages. In the first phase, coding schemes have been developed for conducting three categorical analyses: (1) categorization of tweets, (2) categorization of image use, and (3) categorization of users. The three data sets were then coded using the coding schemes. In the second phase, comparison analyses were conducted among influential, favorited, and general tweets in terms of tweet type, image use, and user. While messages expressing opinion were found to be most favorited, the messages that shared information were recognized as most influential to users. On the other hand, as only four image uses - information dissemination, illustration, emotive/persuasive, and information processing - were found in this data set, the primary image use is likely to be data-driven rather than object-driven. From the perspective of users, the user types such as government, celebrity, and photo-sharing sites were found to be favorited and influential. An improved understanding of how users' image needs, in the context of social media, contribute to the body of knowledge of image needs. This study will also provide valuable insight into practical designs and implications of image retrieval systems or services.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2019.01a
• /
• pp.123-124
• /
• 2019

This paper proposes an improved deep learning method based on small data sets for animal image classification. Firstly, we use a CNN to build a training model for small data sets, and use data augmentation to expand the data samples of the training set. Secondly, using the pre-trained network on large-scale datasets, such as VGG16, the bottleneck features in the small dataset are extracted and to be stored in two NumPy files as new training datasets and test datasets. Finally, training a fully connected network with the new datasets. In this paper, we use Kaggle famous Dogs vs Cats dataset as the experimental dataset, which is a two-category classification dataset.

• The Transactions of the Korea Information Processing Society
• /
• v.7 no.2
• /
• pp.542-551
• /
• 2000

In this paper, a novel technique is presented for automatic brain region segmentation in single channel MR image data sets for 3D visualization and analysis. The method detects brain contours in 2D and 3D processing of four steps. The first and the second make a head mask and an initial brain mask by automatic thresholding using a curve fitting technique. The stage 3 reconstructs 3D volume of the initial brain mask by cubic interpolation and generates an intermediate brain mask using morphological operation and labeling of connected components. In the final step, the brain mask is refined by automatic thresholding using curve fitting. This algorithm is useful for fully automatic brain region segmentation of T1-weighted, T2-weighted, PD-weighted, SPGR MRI data sets without considering slice direction and covering a whole volume of a brain. In the experiments, the algorithm was applied to 20 sets of MR images and showed over 0.97 in comparison with manual drawing in similarity index.