• Journal of Institute of Convergence Technology
• /
• v.11 no.1
• /
• pp.1-5
• /
• 2021

Deep learning based methods achieve state-of-the-art accuracy, however, they typically rely on supervised training with large labeled datasets. It is known in many medical applications that labeling medical images requires significant expertise and much time, and typical hand-tuned approaches for data augmentation fail to capture the complex variations in such images. This paper proposes a 3D image augmentation method to overcome these difficulties. It allows us to enrich diversity of training data samples that is essential in medical image segmentation tasks, thus reducing the data overfitting problem caused by the fact the scale of medical image dataset is typically smaller. Our numerical experiments demonstrate that the proposed approach provides significant improvements over state-of-the-art methods for 3D medical image segmentation.

• Journal of Advanced Navigation Technology
• /
• v.20 no.5
• /
• pp.496-502
• /
• 2016

In this paper, we proposed and implemented a GUI-based system to recognize and record measured values of digital instruments in the industrial site through image processing. Unlike the existing vehicle license plate recognition system, the measured values of the measuring instrument are displayed on the LCD screen as digital numbers. So, the proposed system considers the decimal point, a negative sign, light reflected by LCD protective glass, and various disturbance factors. We used blob-labeling technique to recognize the numbers displayed on the LCD screen, the recognized number images were determined as certain numbers through the template matching, and recognized values were recorded in the storage device with measurement time. Therefore, the proposed system in this paper would reduce the burden of writing when recording the measured values of the inside/outside diameter or height of the product in the industrial site, so effective and errorless process management in production process is possible by preventing errors in recording measurements when written by hand.

• Journal of the Institute of Convergence Signal Processing
• /
• v.2 no.2
• /
• pp.47-54
• /
• 2001

The main purpose of this study is to propose the algorithm about the extraction of the facial feature. To achieve the above goal, first of all, this study produces binary image for input color image. It calculates area after pixel labeling by variant block-units. Secondly, by contour following, circumference have been calculated. So the proper degree of resemblance about area, circumference, the proper degree of a circle and shape have been calculated using the value of area and circumference. And Third, the algorithm about the methods of extracting parameters which are about the feature of eyes, nose, and mouse using the proper degree of resemblance, general structures and characteristics(symmetrical distance) in face have been accomplished. And then the feature parameters of the front face have been extracted. In this study, twelve facial feature parameters have been extracted by 297 test images taken from 100 people, and 92.93 % of the extracting rate has been shown.

• Journal of the Korea Society of Computer and Information
• /
• v.15 no.5
• /
• pp.33-38
• /
• 2010

In this paper, we proposed an algorithm implementing mouse functions using hand motion and number of fingers which are extracted from an image sequence. The sequence is acquired through a web camera and processed with image processing algorithms. The sequence is first converted from RGB model to YCbCr model to efficiently extract skin area and the extracted area is further processed using labeling, opening, and closing operations to decide the center of a hand. Based on the center position, the number of fingers is decided, which serves as the information to decide and perform a mouse function. Experimental results show that 94.0% of pointer moves and 96.0% of finger extractions are successful, which opens the possibility of further development for a commercial product.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.40 no.3
• /
• pp.217-226
• /
• 2003

A novel fingerprint ridge feature extraction using sequential labeling of thinned fingerprint image is proposed, which is invariant to position translation, scaling, and rotation. the proposed algorithm labels ridges of thinned fingerprint image sequentially using vertical line that goes through fingerprint core point. Then, we extract a feature from each labeled ridge and the extraction process is based on the type fo the ridge and a minutiae ridge angle in the ridge. The feature extracted through this process enables us to find out the kind of various minutiae and minutiae angle. As a result of the experiment using two thinned fingerprint images, we finally confirm that proposed algorithm is not related to position translation, scaling, and rotation.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2010.10a
• /
• pp.196-199
• /
• 2010

The human eyes are important facial landmarks for image normalization due to their relatively constant interocular distance. This paper introduces a novel approach for the eye detection task using optimal segmentation method for eye representation. The method consists of three steps: (1)edge extraction method that can be used to accurately extract eye region from the gray-scale face image, (2)extraction of eye region using labeling method, (3)eye localization based on intensity information. Experimental results show that a correct eye detection rate of 98.9% can be achieved on 2408 FERET images with variations in lighting condition and facial expressions.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.14 no.4
• /
• pp.69-77
• /
• 2018

The malignant melanoma accounts for about 1 to 3% of the total malignant tumor in the West, especially in the US, it is a disease that causes more than 9,000 deaths each year. Generally, skin lesions are difficult to detect the features through photography. In this paper, we propose a computer-aided diagnosis algorithm based on deep learning for classification of malignant melanoma and benign skin tumor in RGB channel skin images. The proposed deep learning model configures the tumor lesion segmentation model and a classification model of malignant melanoma. First, U-Net was used to segment a skin lesion area in the dermoscopic image. We could implement algorithms to classify malignant melanoma and benign tumor using skin lesion image and results of expert's labeling in ResNet. The U-Net model obtained a dice similarity coefficient of 83.45% compared with results of expert's labeling. The classification accuracy of malignant melanoma obtained the 83.06%. As the result, it is expected that the proposed artificial intelligence algorithm will utilize as a computer-aided diagnosis algorithm and help to detect malignant melanoma at an early stage.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2019.05a
• /
• pp.338-340
• /
• 2019

Recently, the speed and size of data accumulation are increasing due to the development of networks. There are many difficulties in classifying these data. One of the difficulties is the difficulty of labeling. Labeling is usually done by people, but it is very difficult for everyone to understand the data in the same way and it is very difficult to label them on the same basis. In order to solve this problem, we implemented GAN to generate new image based on input image and to learn input data indirectly by using it for learning. This suggests that the accuracy of classification can be increased by increasing the number of learning data.

• Annual Conference of KIPS
• /
• 2023.05a
• /
• pp.510-512
• /
• 2023

In order to enhance a model's capability for detecting facial expressions, this research suggests a pipeline that makes use of the GradCAM component. The patching module and the pseudo-labeling module make up the pipeline. The patching component takes the original face image and divides it into four equal parts. These parts are then each input into a 2Dconvolutional layer to produce a feature vector. Each picture segment is assigned a weight token using GradCAM in the pseudo-labeling module, and this token is then merged with the feature vector using principal component analysis. A convolutional neural network based on transfer learning technique is then utilized to extract the deep features. This technique applied on a public dataset MMI and achieved a validation accuracy of 96.06% which is showing the effectiveness of our method.

• Applied Microscopy
• /
• v.46 no.3
• /
• pp.134-139
• /
• 2016

Detailed structural and molecular imaging of intact organs has incurred academic interest because the associated technique is expected to provide innovative information for biological investigation and pathological diagnosis. The conventional methods for volume imaging include reconstruction of images obtained from serially sectioned tissues. This approach requires intense manual work which involves inevitable uncertainty and much time to assemble the whole image of a target organ. Recently, effective tissue clearing techniques including CLARITY and ACT-PRESTO have been reported that enables visualization of molecularly labeled structures within intact organs in three dimensions. The central principle of the methods is transformation of intact tissue into an optically transpicuous and macromolecule permeable state without loss of intrinsic structural integrity. The rapidly evolving protocols enable morphological analysis and molecular labeling of normal and pathological characteristics in large assembled biological systems with single-cell resolution. The deep tissue volume imaging will provide fundamental information about mutual interaction among adjacent structures such as connectivity of neural circuits; meso-connectome and clinically significant structural alterations according to pathologic mechanisms or treatment procedures.