• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.5
• /
• pp.2555-2575
• /
• 2017

This paper presents a dual exposure fusion method for image enhancement. Images taken with a short exposure time usually contain a sharp structure, but they are dark and are prone to be contaminated by noise. In contrast, long-exposure images are bright and noise-free, but usually suffer from blurring artifacts. Thus, we fuse the dual exposures to generate an enhanced image that is well-exposed, noise-free, and blur-free. To this end, we present a new scale-space patch-match method to find correspondences between the short and long exposures so that proper color components can be combined within a proposed dual non-local (DNL) means framework. We also present a residual filtering method that eliminates the structure component in the estimated noise image in order to obtain a sharper and further enhanced image. To this end, the entropy is utilized to determine the proper size of the filtering window. Experimental results show that our method generates ghost-free, noise-free, and blur-free enhanced images from the short and long exposure pairs for various dynamic scenes.

• The KIPS Transactions:PartB
• /
• v.18B no.2
• /
• pp.51-56
• /
• 2011

Algorithms for intrinsic images reduce color differences in RGB images caused by the temperature of black-body radiators. Based on the reference light and detecting single invariant direction, these algorithms are weak in real images which can have multiple invariant directions when the scene illuminant is a colored illuminant. To solve these problems, this paper proposes a method of acquiring an intrinsic image by omnidirectional projection of an ROI and a translation of white patch in the ${\chi}$-chromaticity space. Because it is not easy to analyze an image in the three-dimensional RGB space, the ${\chi}$-chromaticity is also employed without the brightness factor in this paper. After the effect of the colored illuminant is decreased by a translation of white patch, an invariant direction is detected by omnidirectional projection of an ROI in this chromaticity space. In case the RGB image has multiple invariant directions, only one ROI is selected with the bin, which has the highest frequency in 3D histogram. And then the two operations, projection and inverse transformation, make intrinsic image acquired. In the experiments, test images were four datasets presented by Ebner and evaluation methods was the follows: standard deviation of the invariant direction, the constancy measure, the color space measure and the color constancy measure. The experimental results showed that the proposed method had lower standard deviation than the entropy, that its performance was two times higher than the compared algorithm.

• Proceedings of the KSRS Conference
• /
• 2002.10a
• /
• pp.708-708
• /
• 2002

Recent developments in sensor technologies have provided remotely sensed data with very high spatial resolution. In order to fully utilize the potential of high resolution images, new image classification strategies are necessary. Unfortunately, the high resolution images increase the spectral within-field variability, and the classification accuracy of traditional methods based on pixel-based classification algorithms such as Maximum-Likelihood method may be decreased (Schiewe 2001). Recent development in Object Oriented Classification based on image segmentation algorithms can be used for the classification of forest patches on rugged terrain of Korea. The objectives of this paper are as follows. First, to compare the pros and cons of image classification methods based on pixel-based and object oriented classification algorithm for the forest patch classification. Landsat ETM＋ data and IKONOS data will be used for the classification. Second, to investigate ways to increase classification accuracy of forest patches. Supplemental data such as DTM and Forest Type Map of 1:25,000 scale are used for topographic correction and image segmentation. Third, to propose the best classification strategy for forest patch classification in terms of accuracy and data requirement. The research site for this paper is Namhansansung Provincial Park located at the eastern suburb of Seoul Metropolitan City for its diverse forest patch types and data availability. Both Landsat ETM＋ and IKONOS data are used for the classification. Preliminary results can be summarized as follows. First, topographic correction of reflectance is essential for the classification of forest patches on rugged terrain. Second, object oriented classification of IKONOS data enables higher classification accuracy compared to Landsat ETM＋ and pixel-based classification. Third, multi-stage segmentation is very useful to investigate landscape ecological aspect of forest communities of Korea.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.11a
• /
• pp.779-782
• /
• 2021

Survival time analysis is one of the main methods used by the pathologist to prognosis for cancer patients. In this paper, we strive to estimate the individual survival time of Adenocarcinoma (ADC) lung cancer patients from pathological images by adopting the convolutional neural network called the SurvPatchV1 model. First, we extracted tissue patches from the whole-slide images (WSI) to deal with extremely large dimensions of WSI. Then the survival time of each patch is estimated through the SurvPatchV1 model. Finally, the individual survival time of each patient is computed. The proposed method is trained and tested on the subset of the NLST dataset for ADC lung cancer. The result demonstrates that our model can obtain all tissue information in lieu of only tumor information in a whole pathological image to estimate the individual survival time.

• Journal of Institute of Control, Robotics and Systems
• /
• v.16 no.12
• /
• pp.1226-1232
• /
• 2010

A door can serve as a feature for place classification and localization for navigation of a mobile robot in indoor environments. This paper proposes a door detection method based on the recognition of various door handles using the general Hough transform (GHT) and support vector machine (SVM). The contour and color histogram of a door handle extracted from the database are used in GHT and SVM, respectively. The door recognition scheme consists of four steps. The first step determines the region of interest (ROI) images defined by the color information and the environment around the door handle for stable recognition. In the second step, the door handle is recognized using the GHT method from the ROI image and the image patches are extracted from the position of the recognized door handle. In the third step, the extracted patch is classified whether it is the image patch of a door handle or not using the SVM classifier. The door position is probabilistically determined by the recognized door handle. Experimental results show that the proposed method can recognize various door handles and detect doors in a robust manner.

• Journal of information and communication convergence engineering
• /
• v.13 no.3
• /
• pp.205-214
• /
• 2015

In this paper, we propose a new approach to learning a discriminative model for figure/ground segmentation by incorporating the bag-of-features and conditional random field (CRF) techniques. We advocate the use of image patches instead of superpixels as the basic processing unit. The latter has a homogeneous appearance and adheres to object boundaries, while an image patch often contains more discriminative information (e.g., local image structure) to distinguish its categories. We use pixel-level sparse coding to represent an image patch. With the proposed feature representation, the unary classifier achieves a considerable binary segmentation performance. Further, we integrate unary and pairwise potentials into the CRF model to refine the segmentation results. The pairwise potentials include color and texture potentials with neighborhood interactions, and an edge potential. High segmentation accuracy is demonstrated on three benchmark datasets: the Weizmann horse dataset, the VOC2006 cow dataset, and the MSRC multiclass dataset. Extensive experiments show that the proposed approach performs favorably against the state-of-the-art approaches.

• The Journal of the Acoustical Society of Korea
• /
• v.43 no.2
• /
• pp.234-242
• /
• 2024

We propose a passive sonar signal classification algorithm using Graph Neural Network (GNN). The proposed algorithm segments spectrograms into image patches and represents graphs through connections between adjacent image patches. Subsequently, Graph Convolutional Network (GCN) is trained using the represented graphs to classify signals. In experiments with publicly available underwater acoustic data, the proposed algorithm represents the line frequency features of spectrograms in graph form, achieving an impressive classification accuracy of 92.50 %. This result demonstrates a 8.15 % higher classification accuracy compared to conventional Convolutional Neural Network (CNN).

• Journal of Broadcast Engineering
• /
• v.13 no.1
• /
• pp.109-118
• /
• 2008

In mapping a texture image into a 3D mesh model for photo-realistic graphic applications, the compression of texture image is as important as geometry of 3D mesh. Typically, the size of the (compressed) texture image of 3D model is comparable to that of the (compressed) 3D mesh geometry. Most 3D model compression techniques are to compress the 3D mesh geometry, rather than to compress the texture image. Well-known image compression standards (i.e., JPEG) have been extensively used for texture image compression. However, such techniques are not so efficient when it comes to compress an image with texture patches, since the patches are little correlated. In this paper, we proposed a preprocessing method to substantially improve the compression efficiency of texture compression. From the experimental results, the proposed method was shown to be efficient in compression with a bit-saving from 23% to 45%.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.16 no.6
• /
• pp.1303-1310
• /
• 2021

We propose an edge-preserving transmission estimation by comparing the patch-based dark channel and the pixel-based dark channel near the edge, in order to improve the quality of outdoor images deteriorated by conditions such as fog and smog. Moreover, we propose a refinement that applies the Guided Image Filtering (GIF), a kind of edge-preserving smoothing filtering methods, to edges using Laplacian operation for natural restoration of image objects and backgrounds, so that we can dehaze a single image and improve the visibility effectively. Experimental results carried out on various outdoor hazy images that show the proposed method has less computational complexity than the conventional methods, while reducing distortion such as halo effect, and showing excellent dehazing performance. In It can be confirmed that the proposed method can be applied to various fields including devices requiring real-time performance.

• Journal of the Korean Graphic Arts Communication Society
• /
• v.23 no.2
• /
• pp.45-58
• /
• 2005

The propose of this study investigated the reproduction of color image displayed on a CRT monitor, for a range of different values of monitor gamma. We have used the GOG(gain-offset-gamma) model of the behavior of the CRT. Color difference have been computed in a color space, based on the CIELAB color appearance model. The 133 patch defined linearly color sample and 24 patch defined printing color target were used, and were subjected to the influence of nine different gamma value. The result show that neutral color is increasing the decrease range of luminance black color than white color. These results are of concern in the context of the "correct" display of color reproduction.