• IEMEK Journal of Embedded Systems and Applications
• /
• v.16 no.5
• /
• pp.209-214
• /
• 2021

Panoptic segmentation, which is now widely used in computer vision such as medical image analysis, and autonomous driving, helps understanding an image with holistic view. It identifies each pixel by assigning a unique class ID, and an instance ID. Specifically, it can classify 'thing' from 'stuff', and provide pixel-wise results of semantic prediction and object detection. As a result, it can solve both semantic segmentation and instance segmentation tasks through a unified single model, producing two different contexts for two segmentation tasks. Semantic segmentation task focuses on how to obtain multi-scale features from large receptive field, without losing low-level features. On the other hand, instance segmentation task focuses on how to separate 'thing' from 'stuff' and how to produce the representation of detected objects. With the advances of both segmentation techniques, several panoptic segmentation models have been proposed. Many researchers try to solve discrepancy problems between results of two segmentation branches that can be caused on the boundary of the object. In this survey paper, we will introduce the concept of panoptic segmentation, categorize the existing method into two representative methods and explain how it is operated on two methods: top-down method and bottom-up method. Then, we will analyze the performance of various methods with experimental results.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.11
• /
• pp.4426-4442
• /
• 2020

Since scene segmentation is becoming a hot topic in the field of autonomous driving and medical image analysis, researchers are actively trying new methods to improve segmentation accuracy. At present, the main issues in image semantic segmentation are intra-class inconsistency and inter-class indistinction. From our analysis, the lack of global information as well as macroscopic discrimination on the object are the two main reasons. In this paper, we propose a Densely connected residual Attention network (DA-Res2Net) which consists of a dense residual network and channel attention guidance module to deal with these problems and improve the accuracy of image segmentation. Specifically, in order to make the extracted features equipped with stronger multi-scale characteristics, a densely connected residual network is proposed as a feature extractor. Furthermore, to improve the representativeness of each channel feature, we design a Channel-Attention-Guide module to make the model focusing on the high-level semantic features and low-level location features simultaneously. Experimental results show that the method achieves significant performance on various datasets. Compared to other state-of-the-art methods, the proposed method reaches the mean IOU accuracy of 83.2% on PASCAL VOC 2012 and 79.7% on Cityscapes dataset, respectively.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.6 no.4
• /
• pp.1188-1202
• /
• 2012

In this paper, we propose a new multi-scale connected coherence tree algorithm (MCCTA) by improving the connected coherence tree algorithm (CCTA). In contrast to many multi-scale image processing algorithms, MCCTA works on multiple scales space of an image and can adaptively change the parameters to capture the coarse and fine level details. Furthermore, we design a Multi-scale Connected Coherence Tree algorithm plus Spectral graph partitioning (MCCTSGP) by combining MCCTA and Spectral graph partitioning in to a new framework. Specifically, the graph nodes are the regions produced by CCTA and the image pixels, and the weights are the affinities between nodes. Then we run a spectral graph partitioning algorithm to partition on the graph which can consider the information both from pixels and regions to improve the quality of segments for providing image segmentation. The experimental results on Berkeley image database demonstrate the accuracy of our algorithm as compared to existing popular methods.

• Smart Structures and Systems
• /
• v.31 no.4
• /
• pp.383-392
• /
• 2023

Post-earthquake building condition assessment is crucial for subsequent rescue and remediation and can be automated by emerging computer vision and deep learning technologies. This study is based on an endeavour for the 2nd International Competition of Structural Health Monitoring (IC-SHM 2021). The task package includes five image segmentation objectives - defects (crack/spall/rebar exposure), structural component, and damage state. The structural component and damage state tasks are identified as the priority that can form actionable decisions. A multi-task Convolutional Neural Network (CNN) is proposed to conduct the two major tasks simultaneously. The rest 3 sub-tasks (spall/crack/rebar exposure) were incorporated as auxiliary tasks. By synchronously learning defect information (spall/crack/rebar exposure), the multi-task CNN model outperforms the counterpart single-task models in recognizing structural components and estimating damage states. Particularly, the pixel-level damage state estimation witnesses a mIoU (mean intersection over union) improvement from 0.5855 to 0.6374. For the defect detection tasks, rebar exposure is omitted due to the extremely biased sample distribution. The segmentations of crack and spall are automated by single-task U-Net but with extra efforts to resample the provided data. The segmentation of small objects (spall and crack) benefits from the resampling method, with a substantial IoU increment of nearly 10%.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.45 no.1
• /
• pp.11-24
• /
• 2008

Region-based image retrieval techniques, which divide image into similar regions having similar characteristics and examine similarities among divided regions, were proposed to support an efficient low-dimensional color indexing scheme. However, color image segmentation techniques are required additionally. The problem of segmentation is difficult because of a large variety of color and texture. It is known to be difficult to identify image regions containing the same color-texture pattern in natural scenes. In this paper we propose an automatic color image segmentation algorithm. The colors in each image are first quantized to reduce the number of colors. The gray level of image representing the outline edge of image is constructed in terms of Fisher's multi-class linear discriminant on quantized images. The gray level of image is transformed into a binary edge image. The edge showing the outline of the binary edge image links to the nearest edge if disconnected. Finally, the final segmentation image is obtained by merging similar regions. In this paper we design and implement a region-based image retrieval system using the proposed segmentation. A variety of experiments show that the proposed segmentation scheme provides good segmentation results on a variety of images.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37 no.2C
• /
• pp.181-187
• /
• 2012

Millimeter wave (MMW) readily penetrates fabrics, thus it can be used to detect objects concealed under clothing. A passive MMW imaging system can operate as a stand-off type sensor that scans people in both indoors and outdoors. However, because of the diffraction limit and low signal level, the imaging system often suffers from low image quality. Therefore, suitable statistical analysis and computational processing would be required for automatic analysis of the images. In this paper, a real-time concealed object detection is addressed by means of the multi-level segmentation. The histogram of the image is modeled with a Gaussian mixture distribution, and hidden object areas are segmented by a multi-level scheme involving $k$-means, the expectation-maximization algorithm, and a decision rule. The complete algorithm has been implemented in C++ environments on a standard computer for a real-time process. Experimental and simulation results confirm that the implemented system can achieve the real-time detection of concealed objects.

• The Transactions of the Korea Information Processing Society
• /
• v.6 no.10
• /
• pp.2809-2815
• /
• 1999

Level set, and active contour(snakes) models are extensively used for image segmentation or shape extraction in computer vision. Snakes utilize the energy minimization concepts, and level set is based on the curve evolution in order to extract contours from image data. In general, these two models have their own drawbacks. For instance, snake acts pooly unless it is placed close to the wanted shape boundary, and it has difficult problem when image has multiple objects to be extracted. But, level set method is free of initial curve position problem, and has ability to handle topology of multiple objects. Nevertheless, level set method requires much more calculation time compared to snake model. In this paper, we use good points of two described models and also apply multi resolution algorithm in order to speed up the process without decreasing the performance of the shape extraction.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2006.05a
• /
• pp.784-787
• /
• 2006

In this paper, we propose a method for nonlinear grapheme segmentation in Korean printed character type classification. The characters are subdivided into six types based on character type information. The feature vector is consist of mesh features, vertical projection features and horizontal projection features which are extracted from gray-level images. We classify characters into 6 types using Back propagation. Character segmentation regions are determined based on character type information. Then, an optimal nonlinear grapheme segmentation path is found using multi-stage graph search algorithm. As the result, a proposed methodology is proper to classify character type and to find nonlinear char segmentation paths.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.58 no.6
• /
• pp.1207-1214
• /
• 2009

In this study, we propose a novel color-image segmentation algorithm to discern the teeth region utilizing RG intensity and its relevant RGB histogram features with resolving the variations of its maximum intensity in terms of peaks and valleys. Tooth candidates in a CCD image are first extracted by applying RGB color multi-threshold levels and consequently the successive morphological image operations and a Sobel-mask edge processing are performed to resolve the teeth region and its contour.

• Journal of Forest and Environmental Science
• /
• v.40 no.1
• /
• pp.15-23
• /
• 2024

This research aimed to assess the possibility of detecting forest degradation using time-series satellite imagery and three different deep learning-based change detection techniques. The dataset used for the deep learning models was composed of two sets, one based on surface reflectance (SR) spectral information from satellite imagery, combined with Texture Information (GLCM; Gray-Level Co-occurrence Matrix) and terrain information. The deep learning models employed for land cover change detection included image differencing using the Unet semantic segmentation model, multi-encoder Unet model, and multi-encoder Unet++ model. The study found that there was no significant difference in accuracy between the deep learning models for forest degradation detection. Both training and validation accuracies were approx-imately 89% and 92%, respectively. Among the three deep learning models, the multi-encoder Unet model showed the most efficient analysis time and comparable accuracy. Moreover, models that incorporated both texture and gradient information in addition to spectral information were found to have a higher classification accuracy compared to models that used only spectral information. Overall, the accuracy of forest degradation extraction was outstanding, achieving 98%.