• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2006년도 Proceedings of ISRS 2006 PORSEC Volume I
• /
• pp.282-285
• /
• 2006

This paper analyzed the characteristics of the Lossy JPEG of the meteorological satellite image, and analyzed the quality of the Lossy JPEG compression, which is proper for the LRIT(Low Rate Information Transmission) to be serviced to the SDUS(Small-scale Data Utilization Station) system of the COMS(Communication, Oceans, Meteorological Satellite). Since COMS is to start running after 2008, we collected the data of the MTSAT-1R(Multi-functional Transport Satellite -1R) for analysis, and after forming the original image to be used to LRIT by each channel and time zone of the satellite image data, we set the different quality with the Lossy JPEG compression, and compressed the original data. For the characteristic analysis of the Lossy JPEG, we measured PSNR(Peak Signal to Noise Rate), compression rate and the time spent in compression following each quality of Lossy JPEG compression. As a result of the analysis of the satellite image data of the MTSAT-1R, the ideal quality of the Lossy JPEG compression was found to be 90% in the VIS Channel, 85% in the IR1 Channel, 80% in the IR2 Channel, 90% in the IR3 Channel and 90% in the IR4 Channel.

• 대한전기학회논문지:시스템및제어부문D
• /
• 제55권3호
• /
• pp.124-130
• /
• 2006

A wavelet-based image compression system allowing both lossless and lossy image compression is proposed in this paper. The proposed algorithm consists of the two stages. The first stage uses the wavelet packet transform and the quad-tree coding scheme for the lossy compression. In the second stage, the residue image taken between the original image and the lossy reconstruction image is coded for the lossless image compression by using the integer wavelet transform and the context based predictive technique with feedback error. The proposed wavelet-based algorithm, allowing an optional lossless reconstruction of a given image, transmits progressively image materials and chooses an appropriate wavelet filter in each stage. The lossy compression result of the proposed algorithm improves up to the maximum 1 dB PSNR performance of the high frequency image, compared to that of JPEG-2000 algorithm and that of S+P algorithm. In addition, the lossless compression result of the proposed algorithm improves up to the maximum 0.39 compression rates of the high frequency image, compared to that of the existing algorithm.

• 한국인쇄학회:학술대회논문집
• /
• 한국인쇄학회 2000년도 춘계 학술발표회 논문집
• /
• pp.17-24
• /
• 2000

The Joint Photographic Experts Group (JPEG) is a standara still-image compression technique, established by the International for Standardization (ISO) and International Telecommunication Standardization Sector (ITUT). The standard is intended to be utilized in the various kinds of color still imaging systems as a standard color image coding format. Because JPEG is a lossy compression, the decompressed image pixel values are nto the same as values before compression. Image of JPEG compression is often made to JPEG recompression at saving to apply JPEG compression of color image. In general, JPEG is a lossy compression and compression image is predicted to be varied image quality according to recompressed Q-factor. Various distortions of JPEG compression and JPEG recompression has been reported in previous paper. In this paper, we compress four difference color samples (photo image, gradient image, vector drawing image, text image) according to various Q-factor, and then compressed images are recompressed according to various Q-factor once again. As the results, we inspect variation of quality and file size of recompressed color image, and ensure the optimum recompression factor.

• 한국인쇄학회지
• /
• 제18권2호
• /
• pp.55-68
• /
• 2000

Joint photographic experts group (JPEG) is a standard still-image compression technique, established by the international organization for standardization (ISO) and international telecommunication standardization sector (ITUT). The standard is intended to be utilized in the various kinds of color still imaging systems as a standard color image coding format. Because JPEG is a lossy compression, the decompressed image pixel values are not the same as the value before compression. Various distortions of JPEG compression and JPEG recompression has been reported in various papers. The Image compressed by JPEG is often recompressed by same type compression method in JPEG. In general, JPEG is a lossy compression and the quality of compressed image is predicted that is varied in according to recompression Q-factor. In this paper, four difference color samples(photo image, gradient image, gradient image, vector drawing image, text image) were compressed in according to various Q-factor, and then the compressed images were recompressed according to various Q-factor once again. As the result, this paper evaluate the variation of image quality and file size in JPEG recompression and recommed the optimum recompression factor.

• 전기학회논문지
• /
• 제58권11호
• /
• pp.2292-2298
• /
• 2009

In this paper, the lossy image compression algorithm using the quad tree and the bandlets is proposed. The proposed algorithm transforms input images by the discrete wavelet transform (DWT) and represents the geometrical structures of high frequency bands using the bandlets with a 8 block- size. In addition, the proposed algorithm searches the position information of the significant coefficients by using the quad tree algorithm and computes the magnitude and the sign information of the significant coefficients by using the Embedded Image Coding using Zerotrees of Wavelet Coefficients (EZW) algorithm. The compression result by using the quad tree algorithm improves the PSNR performance of high frequency images up to 1 dB, compared to that of JPEG-2000 algorithm and that of S+P algorithm. The PSNR performance by using DWT and bandlets improves up to 7.5dB, compared to that by using only DWT.

• 전기학회논문지
• /
• 제56권2호
• /
• pp.415-419
• /
• 2007

Data compression is very important for the storage and transmission of informations. EZW image compression algorithm has been widely used in real application due to its high compression performance. In the EZW algorithm, when a new significant coefficient is generated, its children are all encoded, although its all descendants may be insignificant, and thus its performance is declined. In this paper, we proposed an improved EZW algorithm using IS(Isolated Significant) symbol, which checks all descendants of significant coefficient and avoids encoding the children of each newly generated significant coefficient if it has no significant descendant.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2002년도 ITC-CSCC -2
• /
• pp.912-915
• /
• 2002

In this report, we propose a non-separable two-dimensional (2D) Lossless Wavelet Transform (LWT) for image compression. Filter characteristics of our proposed LWT are the same as those or conventional 2D LWT based on applying 1D LWT twice but our coding performance is better due to reduction of rounding effects. Simulation results confirm effectiveness of our proposed LWT.

• 대한원격탐사학회:학술대회논문집
• /
• 대한원격탐사학회 2002년도 Proceedings of International Symposium on Remote Sensing
• /
• pp.88-88
• /
• 2002

Satellite Image archiving system requires large storage and long transmission time. A simple and cheap way of overcoming these limitations is to increase the compression ratio. However this requires a feasibility study for accurate applications. Here, a new still image compression standard is being developed, the JPEG2000. It provides lossless and lossy compression, progressive transmission by pixel accuracy and by resolution, region-of-interest coding, user-defined tiling size, random codestream access and processing etc. In this study, we will briefly introduce the JPEG2000 compression standard which provides a new compression technique based on the wavelet technology and offers better compression ratios, and evaluate the compression ratios of JPEG2000 for satellite image by performing various image quality tests. Also, we will compare brief test result using the commercial remote sensing software.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2022년도 추계학술대회
• /
• pp.112-115
• /
• 2022

The process of deep learning usually needs to deal with massive data which has greatly limited the development of deep learning technologies today. Convolutional Neural Network (CNN) structure is often used to solve image classification problems. However, a large number of images may be required in order to train an image in CNN, which is a heavy burden for existing computer systems to handle. If the image data can be compressed under the premise that the computer hardware system remains unchanged, it is possible to train more datasets in deep learning. However, image compression usually adopts the form of lossy compression, which will lose part of the image information. If the lost information is key information, it may affect learning performance. In this paper, we will analyze the effect of image compression on deep learning performance on CNN-based cat and dog classification. Through the experiment results, we conclude that the compression of images does not have a significant impact on the accuracy of deep learning.

• 한국멀티미디어학회논문지
• /
• 제25권1호
• /
• pp.29-38
• /
• 2022

Although image compression is one of the essential technologies to transmit image data on a variety of surveillance and mobile healthcare applications, it causes unnecessary compression artifacts such as blocking and ringing artifacts by the lossy compression in the limited network bandwidth. Recently, image restoration methods using convolutional neural network (CNN) show the significant improvement of image quality from the compressed images. In this paper, we propose Image Denoising Convolutional Neural Networks (IDCNN) to reduce the compression artifacts for the purpose of improving the performance of object classification. In order to evaluate the classification accuracy, we used the ImageNet test dataset consisting of 50,000 natural images and measured the classification performance in terms of Top-1 and Top-5 accuracy. Experimental results show that the proposed IDCNN can improve Top-1 and Top-5 accuracy as high as 2.46% and 2.42%, respectively.