• Proceedings of the IEEK Conference
• /
• 대한전자공학회 2006년도 하계종합학술대회
• /
• pp.291-292
• /
• 2006

This paper presents rate-distortion optimization that is considered sequence statistics(complexity) to choose the best macroblock mode decision in H.264. In previous work, Lagrange multiplier is derived by the function of constant value 0.85 and QP so that is not the proper Lagrange multilplier for any image sequence. The proposed algorithm solves the problem by changing constant value 0.85 into adaptive value which is influenced by image complexity, and by reducing the encoder complexity to estimate the image statistics with the multiplication of transformed, quantized rate and distortion. Proposed algorithm is achieved the bit-rate saving up to 5% better than previous method.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• 제57권8호
• /
• pp.1454-1463
• /
• 2008

Blind source separation by independent component analysis (ICA) has applied in signal processing, telecommunication, and image processing to recover unknown original source signals from mutually independent observation signals. Neural networks are learned to estimate the original signals by unsupervised learning algorithm. Because the outputs of the neural networks which yield original source signals are mutually independent, then mutual information is zero. This is equivalent to minimizing the Kullback-Leibler convergence between probability density function and the corresponding factorial distribution of the output in neural networks. In this paper, we present a learning algorithm using information theory and higher order statistics to solve problem of blind source separation. For computer simulation two deterministic signals and a Gaussian noise are used as original source signals. We also test the proposed algorithm by applying it to several discrete images.

• The Korean Journal of Applied Statistics
• /
• 제12권1호
• /
• pp.309-313
• /
• 1999

To overcome current hardship during recent years of university reform, statistics departments of Korean universities should form a new shape with efficient strategies: First, they should value interdisciplinary and open education to foster scientific generalists rather than specialists (Bode.Mosteller.Tukey.Winsor, 1949). Second, they should work out on developing curriculum and improving educational quality for non-statistics majors (Ahn.Cho.Huh, 1994). The service market is widely open and its value is certainly worthy. Third, they may change their department name from "statistics", of which the social image is not quite right, to "data science" or "data information". Statistics is a field of learning on data methodology (Friedman, 1997). methodology (Friedman, 1997).

• Journal of Information Processing Systems
• /
• 제14권2호
• /
• pp.552-562
• /
• 2018

With the advent of the information society, image restoration technology has aroused considerable interest. Guided image filtering is more effective in suppressing noise in homogeneous regions, but its edge-preserving property is poor. As such, the critical part of guided filtering lies in the selection of the guided image. The result of the Expected Patch Log Likelihood (EPLL) method maintains a good structure, but it is easy to produce the ladder effect in homogeneous areas. According to the complementarity of EPLL with guided filtering, we propose a method of coupling EPLL and guided filtering for image de-noising. The EPLL model is adopted to construct the guided image for the guided filtering, which can provide better structural information for the guided filtering. Meanwhile, with the secondary smoothing of guided image filtering in image homogenization areas, we can improve the noise suppression effect in those areas while reducing the ladder effect brought about by the EPLL. The experimental results show that it not only retains the excellent performance of EPLL, but also produces better visual effects and a higher peak signal-to-noise ratio by adopting the proposed method.

• The KIPS Transactions:PartB
• /
• 제13B권6호
• /
• pp.591-600
• /
• 2006

This paper describes the development and implementation of a algorithm to decompose word images from image regions mixed text/graphics in document images using statistical analyses. To decompose word images from image regions, the character components need to be separated from graphic components. For this process, we propose a method to separate them with an analysis of box-plot using a statistics of structural components. An accuracy of this method is not sensitive to the changes of images because the criterion of separation is defined by the statistics of components. And then the character regions are determined by analyzing a local crowdedness of the separated character components. finally, we devide the character regions into text lines and word images using projection profile analysis, gap clustering, special symbol detection, etc. The proposed system could reduce the influence resulted from the changes of images because it uses the criterion based on the statistics of image regions. Also, we made an experiment with the proposed method in document image processing system for keyword spotting and showed the necessity of studying for the proposed method.

• Journal of Korean Society for Quality Management
• /
• 제23권2호
• /
• pp.1-9
• /
• 1995

Two user groups, expert and non-expert, are sampled to measure the image quality of G3 facsimile. A ITU-TS tset chart No. 2 has been transmitted among some selected cities and evaluated by user groups. Their subjective evaluation to the image quality is quantified by Mean Opinion Score method. There is highly significant difference in the image quality between expert and non-expert. From modified logit model, it is concluded that there is no significance in two considered factors, the effects of the number of links and transmission time. The derived percent curves show that 80% of non-experts(90% of expert) is considering the image quality of G3 facsimile "fair, good, or excellent".

• The Journal of Korean Institute of Communications and Information Sciences
• /
• 제38A권2호
• /
• pp.183-190
• /
• 2013

In this paper, we propose a spatially adaptive noise detection and removal algorithm for a single degraded image. Under the assumption that an observed image is Gaussian-distributed, the noise information is estimated by local statistics of degraded image, and the degree of the additive noise is detected by the local statistics of the estimated noise. In addition, we describe a noise removal method taking a modified Gaussian filter which is adaptively determined by filter parameters and window size. The experimental results demonstrate the capability of the proposed algorithm.

• Journal of the Korea Computer Graphics Society
• /
• 제24권3호
• /
• pp.43-48
• /
• 2018

In this paper, we present a method for removing shadows from scanned documents. Compared to the existing methods such as one based on image pyramid representation or adaptive thresholding, our method produces more robust and higher quality results. The basic idea of the approach is to use the local image statistics and to separate interesting regions from the image such as the regions around letters and figures. For the separated regions, we adaptively adjust the local brightness and contrast, and apply the sigmoid function to the intensity values as well to enhance the clarity of the image. For separated the other empty regions, we apply the gradient-base image hole filling method to fill the region with smooth color change.

• The Journal of the Korea institute of electronic communication sciences
• /
• 제9권12호
• /
• pp.1381-1388
• /
• 2014

This paper is proposed a wavelet-based the order statistics MAD(Median Absolute Deviation) method of SAR(Synthetic Aperture Radar) image for image enhancement. also The method of compared and defined the threshold the wavelet coefficients using MAD of the wavelet coefficients of the detail subbands was proposed to effectively image enhancement. In order to complement the disadvantage, the threshold of the proposed method sets up the image statistic and excludes the distortion. The hardware design is used FPGA of Xilinx and DSP system for the image enhancement and compressed encoding of the proposed algorithm. Therefore the proposed method is totally verified by comparing with the several other images.

• Communications for Statistical Applications and Methods
• /
• 제30권3호
• /
• pp.343-354
• /
• 2023

Recently, there have been many improvements in general language models using architectures such as GPT-3 proposed by Brown et al. (2020). Nevertheless, training complex models can hardly be done if the number of data is very small. Data augmentation that addressed this problem was more than normal success in image data. Image augmentation technology significantly improves model performance without any additional data or architectural changes (Perez and Wang, 2017). However, applying this technique to textual data has many challenges because the noise to be added is veiled. Thus, we have developed a novel method for performing data augmentation on text data. We divide the data into signals with positive or negative meaning and noise without them, and then perform data augmentation using k-doc augmentation to randomly combine signals and noises from all data to generate new data.