• Journal of the Korean Institute of Telematics and Electronics
• /
• v.26 no.8
• /
• pp.1238-1246
• /
• 1989

In this paper, a texture synthesis technique based on the NCAR(non-causal auto-regressive) model and the pyramid structure is proposed. In order to estimate the NCAR model parameters accurately from a noisy texture, the MAP(maximum a posteriori) estimation technique is also employed. In our approach, since the input texture is decomposed into the Laplacian oyramid planes first and then the NCAR model is applied to each plane, we are able to obtain a good synthesized texture even if the texture exhibits some non-random local structure or non-homogenity. The usrfulness of the proposed method is demonstrated with seveal real textures in the Brodatz album. Finally, the 2-dimensional MAP estimation technique can be used to the image restoration for noisy images as well as a texture image synthesis.

• Journal of KIISE:Databases
• /
• v.36 no.6
• /
• pp.446-454
• /
• 2009

Among various region detector and shape feature extraction method, MSER(Maximally Stable Extremal Region) and SIFT and its variant methods are popularly used in computer vision application. However, since SIFT is sensitive to the illumination change and MSER is sensitive to the scale change, it is not easy to apply the image similarity retrieval. In this paper, we present a Scale and Rotation Invariant Region Feature(SRIRF) descriptor using scale pyramid, MSER and affine normalization. The proposed SRIRF method is robust to scale, rotation, illumination change of image since it uses the affine normalization and the scale pyramid. We have tested the SRIRF method on various images. Experimental results demonstrate that the retrieval performance of the SRIRF method is about 20%, 38%, 11%, 24% better than those of traditional SIFT, PCA-SIFT, CE-SIFT and SURF, respectively.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.5
• /
• pp.1778-1797
• /
• 2021

Document image binarization is an important pre-processing step in document analysis and archiving. The state-of-the-art models for document image binarization are variants of encoder-decoder architectures, such as FCN (fully convolutional network) and U-Net. Despite their success, they still suffer from three limitations: (1) reduced feature map resolution due to consecutive strided pooling or convolutions, (2) multiple scales of target objects, and (3) reduced localization accuracy due to the built-in invariance of deep convolutional neural networks (DCNNs). To overcome these three challenges, we propose an improved semantic segmentation model, referred to as DP-LinkNet, which adopts the D-LinkNet architecture as its backbone, with the proposed hybrid dilated convolution (HDC) and spatial pyramid pooling (SPP) modules between the encoder and the decoder. Extensive experiments are conducted on recent document image binarization competition (DIBCO) and handwritten document image binarization competition (H-DIBCO) benchmark datasets. Results show that our proposed DP-LinkNet outperforms other state-of-the-art techniques by a large margin. Our implementation and the pre-trained models are available at https://github.com/beargolden/DP-LinkNet.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 1995.06a
• /
• pp.97-100
• /
• 1995

In transmission of image sequences over ATM networks, the packet loss problem and channel sharing efficiency are important. As a possible solution two-layer video coding methods have been proposed. These methods transmit video information over the network with different levels of protection with respect to packets loss. In this paper, a two-layer coding method using pyramid structure is proposed and several realizations of two-layer video coding methods are presented and their performances are compared.

• Journal of Korean Society for Geospatial Information Science
• /
• v.20 no.3
• /
• pp.41-48
• /
• 2012

Using multi-sensor or multi-temporal high resolution satellite images together is essential for efficient applications in remote sensing area. The purpose of this paper is to estimate geometric difference of translations between high-resolution optical and SAR images automatically. The geometric and radiometric pre-processing steps were fulfilled to calculate the similarity between optical and SAR images by using Mutual Information method. The coarsest-level pyramid images of each sensor constructed by gaussian pyramid method were generated to estimate the initial translation difference of the x, y directions for calculation efficiency. The precise geometric difference of translations was able to be estimated by applying this method from coarsest-level pyramid image to original image in order. Yet even when considered only translation between optical and SAR images, the proposed method showed RMSE lower than 5m in all study sites.

• The Transactions of the Korea Information Processing Society
• /
• v.3 no.6
• /
• pp.1624-1635
• /
• 1996

Palletized color images are the dominant type of the image used on Internet and World-Wide Web. In spite of this, most image compression and progressive transmission algorithm have been designed for continuous-tone images. Pallettize images. Palletized images differ from continuous-tone images in such a aspect that values are lookup table indices a new pyramid structure for compression and progressive transmission of a palletized image. In the proposed pyramid structure, the color of a node at higher level is the one that occupies the most part in 4 sons and each node is represented by a type code and several color codes. Since the proposed method do not exploit spatial correlation in an image, it is ideally applied to lossless compression and progressive transmission of palletized images. We have confirmed this through the experimental results.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.5
• /
• pp.1814-1828
• /
• 2021

Low-light image enhancement is a key technique to overcome the quality degradation of photos taken under scotopic vision illumination conditions. The degradation includes low brightness, low contrast, and outstanding noise, which would seriously affect the vision of the human eye recognition ability and subsequent image processing. In this paper, we propose an approach based on deep learning and Retinex theory to enhance the low-light image, which includes image decomposition, illumination prediction, image reconstruction, and image optimization. The first three parts can reconstruct the enhanced image that suffers from low-resolution. To reduce the noise of the enhanced image and improve the image quality, a super-resolution algorithm based on the Laplacian pyramid network is introduced to optimize the image. The Laplacian pyramid network can improve the resolution of the enhanced image through multiple feature extraction and deconvolution operations. Furthermore, a combination loss function is explored in the network training stage to improve the efficiency of the algorithm. Extensive experiments and comprehensive evaluations demonstrate the strength of the proposed method, the result is closer to the real-world scene in lightness, color, and details. Besides, experiments also demonstrate that the proposed method with the single low-light image can achieve the same effect as multi-exposure image fusion algorithm and no ghost is introduced.

• Journal of the Semiconductor & Display Technology
• /
• v.19 no.2
• /
• pp.11-15
• /
• 2020

In a typical initial setup of a PCB component inspection system, operators should manually input various information such as category, position, and inspection area for each component to be inspected, thus causing much inconvenience and longer setup time. Although there are many deep learning based object detectors, RetinaNet is regarded as one of best object detectors currently available. In this paper, a method using an extended RetinaNet is proposed that automatically detects its component category and position for each component mounted on PCBs from a high-resolution color input image. We extended the basic RetinaNet feature pyramid network by adding a feature pyramid layer having higher spatial resolution to the basic feature pyramid. It was demonstrated by experiments that the extended RetinaNet can detect successfully very small components that could be missed by the basic RetinaNet. Using the proposed method could enable automatic generation of inspection areas, thus considerably reducing the setup time of PCB component inspection systems.

• The Journal of the Korea Contents Association
• /
• v.21 no.3
• /
• pp.1-9
• /
• 2021

The hologram pyramid is an application of floating holograms, allowing the observer to see three-dimensional holograms from various angles without wearing wearable devices. Due to the low cost and ease of manufacturing, it has been used in a wide variety of fields as diverse as education, prototyping, showcase, and etc. But, when the observer looks at the hologram from the place where each side of the hologram pyramid is connected, the hologram looks cut and distorted. Also, the observer can see the only hologram of angles viewed head-on from each side. In this paper, we propose a method of generating a hologram image corresponding to the observer's gaze angle by tracking the observer's position and conducting reverse distortion. It provide a hologram of the angle viewed by the observer without cutting and distortion. In addition, the existing method and the proposed method were applied and compared in the hologram pyramid.

• Proceedings of the KIEE Conference
• /
• 1992.07a
• /
• pp.449-452
• /
• 1992

Before performing an adaptive stereo matching using satellite images, classification is needed as a preprocessing step. This paper describes that classification of three land cover types : river, mountain, and agricultural fields. We proposed that classification algorithm using pyramid structure and texture features. Results of applying the proposed algorithm to satellite image improved classification accuracy.