• Smart Media Journal
• /
• v.9 no.2
• /
• pp.9-15
• /
• 2020

In this paper, we propose a network model composed of Multi path Encoder-Decoder branches that can recognize each instance from the image. The network has two branches, Dot branch and Segmentation branch for finding the center point of each instance and for recognizing area of the instance, respectively. In the experiment, the CVPPP dataset was studied to distinguish leaves from each other, and the center point detection branch(Dot branch) found the center points of each leaf, and the object segmentation branch(Segmentation branch) finally predicted the pixel area of each leaf corresponding to each center point. In the existing segmentation methods, there were problems of finding various sizes and positions of anchor boxes (N > 1k) for checking objects. Also, there were difficulties of estimating the number of undefined instances per image. In the proposed network, an effective method finding instances based on their center points is proposed.

• Journal of Information Technology Applications and Management
• /
• v.23 no.2
• /
• pp.11-28
• /
• 2016

In this paper, we proposed a dictionary extension and a ternary decomposition technique to improve the effectiveness of Khmer word segmentation. Most word segmentation approaches depend on a dictionary. However, the dictionary being used is not fully reliable and cannot cover all the words of the Khmer language. This causes an issue of unknown words or out-of-vocabulary words. Our approach is to extend the original dictionary to be more reliable with new words. In addition, we use ternary decomposition for the segmentation process. In this research, we also introduced the invisible space of the Khmer Unicode (char\u200B) in order to segment our training corpus. With our segmentation algorithm, based on ternary decomposition and invisible space, we can extract new words from our training text and then input the new words into the dictionary. We used an extended wordlist and a segmentation algorithm regardless of the invisible space to test an unannotated text. Our results remarkably outperformed other approaches. We have achieved 88.8%, 91.8% and 90.6% rates of precision, recall and F-measurement.

• Smart Media Journal
• /
• v.2 no.2
• /
• pp.14-19
• /
• 2013

Recently, due to the increase of the use of scanned images, the text segmentation techniques, which play critical role to optimize the quality of the scanned images, are required to be updated and advanced. In this study, an algorithm has been developed based on the modification of Gaussian mixture model (GMM) by integrating the calculation of Gaussian detection gradient and the estimation of the number clusters. The experimental results show an efficient method for text segmentation in natural scenes such as storefronts, street signs, scanned journals and newspapers at different size, shape or color of texts in condition of lighting changes and complex background. These indicate that our model algorithm and research approach can address various issues, which are still limitations of other senior algorithms and methods.

• Smart Media Journal
• /
• v.2 no.2
• /
• pp.8-13
• /
• 2013

Nowadays, the saliency region detection has become a popular research topic because of its uses for many applications like object recognition and object segmentation. Some of recent methods apply color distinctiveness based on an analysis of statistics of color image derivatives in order to boosting color saliency can produce the good saliency maps. However, if the salient regions comprise more than half the pixels of the image or the background is complex, it may cause bad results. In this paper, we introduce the method to handle these problems by using maximum symmetric surround. The results show that our method outperforms the previous algorithms. We also show the segmentation results by using Otsu's method.

• Smart Media Journal
• /
• v.9 no.3
• /
• pp.59-70
• /
• 2020

This paper presents a fully automatic tool to recognize the liver region from CT images based on a deep learning model, namely Multiple Filter U-net, MFUnet. The advantages of both U-net and Multiple Filters were utilized to construct an autoencoder model, called MFUnet for segmenting the liver region from computed tomograph. The MFUnet architecture includes the autoencoding model which is used for regenerating the liver region, the backbone model for extracting features which is trained on ImageNet, and the predicting model used for liver segmentation. The LiTS dataset and Chaos dataset were used for the evaluation of our research. This result shows that the integration of Multiple Filter to U-net improves the performance of liver segmentation and it opens up many research directions in medical imaging processing field.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2021.07a
• /
• pp.81-83
• /
• 2021

4차 산업 혁명과 함께 인공지능이 발전 하고 있다. 그 CNN 등 과 같은 이미지 관련 신경망들이 발전되어 가스 탐지와 같은 여러 분야에서 사용되고 있다. 하지만 가스 탐지는 Box 형태의 탐지가 일반적이고 Segmentation에 관한 연구는 있지만 연기와 같이 경계선이 불분명한 개체에 대해서는 연구가 미비하다. 본 논문에서는 Segmentation에 강력한 성능을 보이는 U-net을 활용하여 Box 형태가 아닌 Segmentation을 진행하여 픽셀단위로 연기를 탐지하고자 한다.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.13 no.2
• /
• pp.233-237
• /
• 2013

Image recognition is an important research area of pattern recognition. This paper studies that the image segmentation algorithm theory and its application in road signs recognition system. In this paper We studied a systematic study for road signs and we have made the recognition algorithm. This paper is divided in image segmentation part and image recognition part for the road signs recognition. The experimental results show that the road signs recognition model can make effective use in smart phone system, and the model can be used in many other fields.

• Journal of the Korean Society of Radiology
• /
• v.18 no.3
• /
• pp.213-222
• /
• 2024

This study analyzed the volumes generated by deep learning and atlas-based automatic segmentation methods, as well as the Dice similarity coefficient and 95% Hausdorff distance, according to the conditions of conduction voltage and conduction current in computed tomography for lung radiotherapy. The first result, the volumes generated by the atlas-based smart segmentation method showed the smallest volume change as a function of the change in tube voltage and tube current, while Aview RT ACS and OncoStudio using deep learning showed smaller volumes at tube currents lower than 100 mA. The second result, the Dice similarity coefficient, showed that Aview RT ACS was 2% higher than OncoStuido, and the 95% Hausdorff distance results also showed that Aview RT ACS analyzed an average of 0.2-0.5% higher than OncoStudio. However, the standard deviation of the respective results for tube current and tube voltage is lower for OncoStudio, which suggests that the results are consistent across volume variations. Therefore, caution should be exercised when using deep learning-based automatic segmentation programs at low perfusion voltages and low perfusion currents in CT imaging conditions for lung radiotherapy, and similar results were obtained with conventional atlas-based automatic segmentation programs at certain perfusion voltages and perfusion currents.

• Smart Media Journal
• /
• v.1 no.2
• /
• pp.34-39
• /
• 2012

In this paper, post-processing step is presented to remove noises and develop a fully automated scheme to segment flowers in natural scene images. The scheme to segment flowers using a level set algorithm in the natural scene images produced unexpected and isolated noises because the level set relies only on the color and edge information. The experimental results shows that the proposed method successfully removes noises in the foreground and background.

• International journal of advanced smart convergence
• /
• v.11 no.4
• /
• pp.81-87
• /
• 2022

In this paper, we propose a watershed algorithm that applies a high-frequency enhancement filter to emphasize the boundary and a local adaptive threshold to search for minimum points. The previous method causes the problem of over-segmentation, and over- segmentation appears around the boundary of the object, creating an inaccurate boundary of the region. The proposed method applies a high-frequency enhancement filter that emphasizes the high-frequency region while preserving the low-frequency region, and performs a minimum point search to consider local characteristics. When merging regions, a fixed threshold is applied. As a result of the experiment, the proposed method reduced the number of segmented regions by about 58% while preserving the boundaries of the regions compared to when high frequency emphasis filters were not used.