• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.7
• /
• pp.1706-1725
• /
• 2024

The unique U-shaped structure of U-Net network makes it achieve good performance in image segmentation. This network is a lightweight network with a small number of parameters for small image segmentation datasets. However, when the medical image to be segmented contains a lot of detailed information, the segmentation results cannot fully meet the actual requirements. In order to achieve higher accuracy of medical image segmentation, a novel improved U-Net network architecture called multi-scale encoder-decoder U-Net+ (MEDU-Net+) is proposed in this paper. We design the GoogLeNet for achieving more information at the encoder of the proposed MEDU-Net+, and present the multi-scale feature extraction for fusing semantic information of different scales in the encoder and decoder. Meanwhile, we also introduce the layer-by-layer skip connection to connect the information of each layer, so that there is no need to encode the last layer and return the information. The proposed MEDU-Net+ divides the unknown depth network into each part of deconvolution layer to replace the direct connection of the encoder and decoder in U-Net. In addition, a new combined loss function is proposed to extract more edge information by combining the advantages of the generalized dice and the focal loss functions. Finally, we validate our proposed MEDU-Net+ MEDU-Net+ and other classic medical image segmentation networks on three medical image datasets. The experimental results show that our proposed MEDU-Net+ has prominent superior performance compared with other medical image segmentation networks.

• Journal of KIISE:Software and Applications
• /
• v.29 no.11
• /
• pp.809-817
• /
• 2002

This paper describes a new face detection method that is a pre-processing algorithm for on-line face recognition. To complement the weakness of using only edge or rotor features from previous face detection method, we propose the two types of face detection method. The one is a combined method with edge and color features and the other is a center area color sampling method. To prevent connecting the people's face area and the background area, which have same colors, we propose a new adaptive edge detection algorithm firstly. The adaptive edge detection algorithm is robust to illumination variance so that it extracts lots of edges and breakouts edges steadily in border between background and face areas. Because of strong edge detection, face area appears one or multi regions. We can merge these isolated regions using color information and get the final face area as a MBR (Minimum Bounding Rectangle) form. If the size of final face area is under or upper threshold, color sampling method in center area from input image is used to detect new face area. To evaluate the proposed method, we have experimented with 2,100 face images. A high face detection rate of 96.3% has been obtained.

• Journal of Sensor Science and Technology
• /
• v.11 no.5
• /
• pp.255-262
• /
• 2002

Human retina is able to detect the edge of an object effectively. We designed a CMOS vision chip by modeling cells of the retina as hardwares involved in edge detection. There are several fluctuation factors which affect characteristics of MOSFETs during CMOS fabrication process and this effect appears as output offset of the vision chip which is composed of pixel arrays and readout circuits. The vision chip detecting edge information from input image is used for input stage of other systems. Therefore, the output offset of a vision chip determine the efficiency of the entire performance of a system. In order to eliminate the offset at the output stage, we designed a vision chip by using CDS(Correlated Double Sampling) technique. Using standard CMOS process, it is possible to integrate with other circuits. Having reliable output characteristics, this chip can be used at the input stage for many applications, like targe tracking system, fingerprint recognition system, human-friendly robot system and etc.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.19 no.12
• /
• pp.2785-2791
• /
• 2015

In this paper, neddle detection algorithm for the removal of needle stuck into skin in oriental clinic is presented. First, in the proposed method, potential candidate areas of each needle are selected by using the morphological open operation in a gray image, and the false candidates are removed by considering their area size. Next, edge points are extracted using canny edge detector in selected candidate areas, line segments are approximated using the edge points. Based on the direction of line segment and the distance between two line segments, two main line segments of the needle are extracted. The final verification of needle is accomplished by using the morphological analysis of these two line segments. In the experiments, the detection rate of proposed method reaches to 97.5% for the 16 images containing 119 needles.

• Journal of Broadcast Engineering
• /
• v.14 no.2
• /
• pp.253-264
• /
• 2009

This paper presents an algorithm, aims at practical applications, for the high speed processing and performance enhancement of lane detection base on vision processing system. As a preprocessing for high speed lane detection, the vanishing line estimation and the optimal extraction of region of interest for lane boundary (ROI-LB) can be processed to reduction of detection region in which high speed processing is enabled. Image feature information is extracted only in the ROI-LB. Road lane is extracted using a non-parametric model fitting and Hough transform within the ROI-LB. With simultaneous processing of noise reduction and edge enhancement using the Laplacian filter, the reliability of feature extraction can be increased for various road lane patterns. Since outliers of edge at each block can be removed with clustering of edge orientation for each block within the ROI-LB, the performance of lane detection can be greatly improved. The various real road experimental results are presented to evaluate the effectiveness of the proposed method.

• Journal of the Korea Society of Computer and Information
• /
• v.17 no.7
• /
• pp.67-76
• /
• 2012

This paper proposes an efficient method that helps a visually impaired person to detect a pavement borderline. A pedestrian is equipped with a camera so that the front view of a natural scene is captured. Our approach analyzes the captured image and detects the borderline of a pavement in a very robust manner. Our approach performs the task in two steps. In a first step, our approach detects a vanishing point and vanishing lines by applying an edge operator. The edge operator is designed to take a threshold value adaptively so that it can handle a dynamic environment robustly. The second step is to determine the borderlines of a pavement based on vanishing lines detected in the first step. It analyzes the vanishing lines to form VRays that confines the pavement only. The VRays segments out the pavement region in a radial manner. We compared our approach against Canny edge detector. Experimental results show that our approach detects borderlines of a pavement very accurately in various situations.

• Journal of IKEEE
• /
• v.7 no.2 s.13
• /
• pp.197-209
• /
• 2003

A current TV out format is quite different from that of HDTV or PC monitor in encoding techniques. In other words, a conventional analog TV uses interlaced display while HDTV or PC monitor uses Non-interlaced / Progressive-scanned display. In order to encode image signals coming from devices that takes interlaced display format for progressive scanned display, a hardware logic in which scanning and interpolation algorithms are implemented is necessary. The ELA(Edge-Based Line Average) algorithm have been widely used because it provided good characteristics. In this study, the ADI(Adaptive De-interlacing Interpolation) algorithm using to improve the ELA algorithm which shows low quality in vertical edge detections and low efficiency of horizontal edge lines. With the De-interlacing ASIC chip that converts the interlaced Digital YUV to De-interlaced Digital RGB is designed. The VHDL is used for chip design.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.7 no.6
• /
• pp.1100-1105
• /
• 2006

Severe distortions of colors in the obtained images have made it difficult for color codes to expand their applications. To reduce the effect of color distortions on reading colors, it will be more desirable to statistically process as many pixels in the individual color region as possible, than relying on some regularly sampled pixels. This process may require segmentation, which usually requires edge detection. However, edges in color codes can be disconnected due tovarious distortions such as zipper effect and reflection, to name a few, making segmentation incomplete. Edge linking is also a difficult process. In this paper, a more efficient approach to reducing the effect of color distortions on reading colors, one that excludes precise edge detection for segmentation, was obtained by employing the k-means clustering algorithm. And, in detecting color codes, the properties of both six safe colors and grays were utilized. Experiments were conducted on 144, 4M-pixel, outdoor images. The proposed method resulted in a color-code detection rate of 100% fur the test images, and an average color-reading accuracy of over 99% for the detected codes, while the highest accuracy that could be achieved with an approach employing Canny edge detection was 91.28%.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.12
• /
• pp.93-100
• /
• 2022

In this paper, we propose a mobile app for personalized rehabilitation exercise coaching and management service using an edge computing-based personalized exercise information collection system. The existing management method that relies on user input information has difficulty in examining the actual possibility of rehabilitation. In this paper, we implement an application that collects movement information along with body joint information through image information analysis based on edge computing at a remote location, measures the time and accuracy of the movement, and provides rehabilitation progress through correct posture information. In addition, in connection with the measurement equipment of the rehabilitation center, the health status can be managed, and the accuracy of exercise information and trend analysis information is provided. The results of this study will enable management and coaching according to self-rehabilitation exercises in a contactless environment.

• Journal of Biomedical Engineering Research
• /
• v.44 no.2
• /
• pp.147-156
• /
• 2023

Efforts to employ smart home sensors to monitor the indoor activities of elderly single residents have been made to assess the feasibility of a safe and healthy lifestyle. However, the bathroom remains an area of blind spot. In this study, we have developed and evaluated a new edge computer device that can automatically detect water usage activities in the bathroom and record the activity log on a cloud server. Three kinds of sound as flushing, showering, and washing using wash basin generated during water usage were recorded and cut into 1-second scenes. These sound clips were then converted into a 2-dimensional image using MEL-spectrogram. Sound data augmentation techniques were adopted to obtain better learning effect from smaller number of data sets. These techniques, some of which are applied in time domain and others in frequency domain, increased the number of training data set by 30 times. A deep learning model, called CRNN, combining Convolutional Neural Network and Recurrent Neural Network was employed. The edge device was implemented using Raspberry Pi 4 and was equipped with a condenser microphone and amplifier to run the pre-trained model in real-time. The detected activities were recorded as text-based activity logs on a Firebase server. Performance was evaluated in two bathrooms for the three water usage activities, resulting in an accuracy of 96.1% and 88.2%, and F1 Score of 96.1% and 87.8%, respectively. Most of the classification errors were observed in the water sound from washing. In conclusion, this system demonstrates the potential for use in recording the activities as a lifelog of elderly single residents to a cloud server over the long-term.