• Korean Journal of Remote Sensing
• /
• v.11 no.2
• /
• pp.43-57
• /
• 1995

This paper presents reconstruction of disparity in images. To achieve this, the algorithm was made up of two different procedures - one is extraction of boundaries for man-made structures and the other is matching of the structures. In the extraction of boundaries for man-made structures, we assume that man-made structures are composed of lines and the lines make up closed polygon. The convertional algorithms of the edges extraction may not perceive man-made structures and have problems that matching algorithms were too complex. This paper proposed sub-pixel boundaries extraction algorithm that fused split-and-merge and image improvement algorithms to overcome complexity. In matching procedure, feature-based algorithm that minimize the proposed cost function are used and the cost fuction considers movement of mid-points for left and right images to match structures. Because we could not obtain disparity of inner parts for the man-made structures, interpolation method was used. The experiment showed good results.

• Journal of Semiconductor Engineering
• /
• v.1 no.1
• /
• pp.31-37
• /
• 2020

We present an approximate DRAM architecture for energy-efficient deep learning. Our key premise is that by bounding memory errors to non-critical information, we can significantly reduce DRAM refresh energy without compromising recognition accuracy of deep neural networks. To validate the key premise, we make extensive Monte-Carlo simulations for several well-known convolutional neural networks such as LeNet, ConvNet and AlexNet with the input of MINIST, CIFAR-10, and ImageNet, respectively. We assume that the highest-order 8-bits (in single precision) and 4-bits (in half precision) are protected from retention errors under the proposed architecture and then, randomly inject bit-errors to unprotected bits with various bit-error-rates. Here, recognition accuracies of the above convolutional neural networks are successfully maintained up to the 10^-5-order bit-error-rate. We simulate DRAM energy during inference of the above convolutional neural networks, where the proposed architecture shows the possibility of considerable energy saving up to 10 ~ 37.5% of total DRAM energy.

• The Research Journal of the Costume Culture
• /
• v.19 no.2
• /
• pp.389-401
• /
• 2011

Jewelry is not only symbolic meaning of the fortune, but also it completes or redound to fashion style and its image in this modern times which diverse culture live together. The instinct for adornment which is long as human cultures is developed as Artwork genuinely to show effective self-expression that is aesthetic and distinguished. It would be from that it made by using metallic materials. In contemporary fashion, jewelry takes so much importance that it sets the trend. They carry a sensible message that expresses esthetic desire and originality. Among the so many kinds of jewelry, especially the necklace is located beneath the face and linked as a part of fashion, so it frequently has showed in Fashion Collection, We tend to study the effect of that the form, hue and character of materials of necklace that is expressed in fashion collection influence fashion image. The method of this study is comprised with precedent studies and analysis of necklace photos in fashion collection. For the analysis of data, we implement content analysis and statistical analysis using SPAW Statistics 18. As the result, fashion and jewelry effect interactively and share esthetic forms, in the view of total image necklace image is more strong than fashion image. Because the hue and the form of necklace take a great role to make fashion image with the sense of its eyesight, its effective coordination go up the delicate feelings of the fashion. So, it is very effective things that we predict the trend of fashion, then, coordinate with well-matched necklace.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.06a
• /
• pp.314-319
• /
• 1998

In this paper, for efficient data classification in multispectral bands environment, a band feature extraction method using the Rough sets theroy is proposed. First, we make a look up table from training data, and analyze the properties of experimental multispectral image data, then select the efficient band usin indiscernibility relation of Rough sets theory from analysis results. Proposed method is applied to LAMDSAT TM data on 2, June, 1992. Among them, normal distributive data were experimented, mainly. From this, we show clustering trends that similar to traditional band selection results by wavelength properties, from this, we verify that can use the proposed method that centered on data properties to select the efficient bands, though data sensing environment change to hyperspectral band environments.

• Journal of Korea Multimedia Society
• /
• v.7 no.5
• /
• pp.640-647
• /
• 2004

This paper presents a method that, by using a tree splay, dynamically modifies the fixed table of a VLC used in the standards and thus improves the performance of a video encoder. The previous standards, which drew up tables by using a limited number of images, were not able to accommodate the characteristics of new image input. The new method proposed in this paper employs the probabilistic characteristics of an image input, thus making encoding more effective. Furthermore, the present study also proposes a method to regulate the weight in order that the accumulated probabilities can make a dominant symbol emerge, while keeping the probability adequate and thus maximally accommodating the characteristics of the image input. The results from applying the proposed method to a variety of image sequences showed that, while various types of images yielded slightly different results, it improved the performance up to a maximum of 0.23㏈, compared to MPEG-4 standard.

• The KIPS Transactions:PartB
• /
• v.15B no.3
• /
• pp.189-196
• /
• 2008

In this paper, a CMOS digital image sensor, which consists of A/D conversion, motion estimation circuits, and an attention module for ROI (Region of Interest) detection is presented. The functions of A/D conversion and motion estimation are implemented by $0.6{\mu}m$ CMOS processing circuit as hardware, and the attention module is implemented outside the circuit as software currently. Attention modules are taken to improve limited applications of the smart image sensor. The current smart image sensor responses to the changes of intensity, and uses the integration time to estimate motion. Therefore it is limited in its applications. To make up for inherent property of the sensor from circuit design and extend its applications we decide to introduce perception solutions to the image sensor. Attention modules for still and moving images are employed to achieve such purposes. The suggested approach makes the smart image sensor available with additional functions for such cases that motion estimation or intensity changes are not observed. Experimental result shows the usefulness and extension of the image sensor.

• Journal of the Korea Society of Computer and Information
• /
• v.18 no.10
• /
• pp.39-46
• /
• 2013

Recently, the mobile industries grow up rapidly and their performances are improved. So the usage of mobile devices is increasing in our life. Also mobile devices equipped with a high-performance camera, so the image stitching can carry out on the mobile devices instead of the desktop. However the mobile devices have limited hardware to perform the image stitching which has a lot of computational complexity. In this paper, we have proposed improved feature descriptor extraction and matching method for efficient image stitching on mobile environment. Our method can reduce computational complexity using extension of orientation window and reduction of dimension feature descriptor when feature descriptor is generated. In addition, the computational complexity of image stitching is reduced through the classification of matching points. In our results, our method makes to improve the computational time of image stitching than the previous method. Therefore our method is suitable for the mobile environment and also that method can make natural-looking stitched image.

• Proceedings of the Korean Printing Society Conference
• /
• 2007.11a
• /
• pp.13-28
• /
• 2007

Producing printing plate is essential progress to do offset printing. In this Film-less period, the more PS plate becomes extinct, the more the age of the Plate-Making of Exposure declines the place to stand. To do offset printing, the CTP (Computer to Plate) is taking a place of PS plate that covers speed, quality and economical problems. The biggest advantage of using CTP is that laser directly goes to the plate, thus there are no dust from the Plate-Making of Exposure. It is also theoretically able to print 200lpi${\sim}$300lpi as well as print 1751pi, because it has over 2400dpi resolution. The high quality printing could be available inside of the country, if printing machine keeps the optimum condition in offset printing. The CTP has many advantages, however there is a difficulty for the operators to preserve the equipment. The actual circumstance is that they cannot make a decision about how many dots need to be generated, and also it is necessary to know how to establish the setup at RIP on CTP to make the optimum condition output. If offset printing machine keeps the optimum condition, it would be able to print up to high quality printing however it is hard to comment what is the optimum condition for the printing machine. Anyone could say easy subjectively that machine is in the optimum condition, however it is objectively hard to estimate by number. In this research GATF / Plate Test target used to analyze the image and to make numerical value of the optimum condition of the CTP. It also used GATF / The sheep fed test printing 5.0 to know the density of the color representation, dot gain and gray balance for the optimum condition of the print machine. The purpose of this research is to represent the ISO 12647-2 which is the international standard with domestic printing equipments.

• Journal of the Korean Graphic Arts Communication Society
• /
• v.25 no.2
• /
• pp.69-84
• /
• 2007

Producing printing plate is essential progress to do offset printing. In this Film-less period, the more PS plate becomes extinct, the more the age of the Plate-Making of Exposure declines the place to stand. To do offset printing, the CTP (Computer to Plate) is taking a place of PS plate that covers speed, quality and economical problems. The biggest advantage of using CTP is that laser directly goes to the plate, thus there are no dust from the Plate-Making of Exposure. It is also theoretically able to print $200lpi{\sim}300lpi$ as well as print 1751pi, because it has over 2400dpi resolution. The high quality printing could be available inside of the country, if printing machine keeps the optimum condition in offset printing. The CTP has many advantages, however there is a difficulty for the operators to preserve the equipment. The actual circumstance is that they cannot make a decision about how many dots need to be generated, and also it is necessary to know how to establish the setup at RIP on CTP to make the optimum condition output. If offset printing machine keeps the optimum condition, it would be able to print up to high quality printing however it is hard to comment what is the optimum condition for the printing machine. Anyone could say easy subjectively that machine is in the optimum condition, however it is objectively hard to estimate by number. In this research GATF / Plate Test target used to analyze the image and to make numerical value of the optimum condition of the CTP. It also used GATF / The sheep fed test printing 5.0 to know the density of the color representation, dot gain and gray balance for the optimum condition of the print machine. The purpose of this research is to represent the ISO 12647-2 which is the international standard with domestic printing equipments.

• Journal of the Korean Society of Costume
• /
• v.60 no.9
• /
• pp.16-25
• /
• 2010

This study reinterpreted the formative design elements of traditional grate patterns to create new lattice patterns and come up with a design concept for fashion-cultural products that highlight the uniqueness of traditional Korean culture and its characteristic features. Methodologically, the computer design software programs Adobe Illustrator CS2 and Adobe Photoshop were used to make grate patterns motifs. and they were applied to scarves and again to blouses using a three-dimensional simulation technique. In this study, three basic motifs for a new formative image were set using graphical functions such as omitting, simplifying, overlapping, repeating and reducing shapes based on the images of traditional 亞-shaped, arched and floral lattices, and each motif was expanded to have two variations with different colors applied to them. The direction of basic motif design was set to fit for each of fashion-cultural items such as scarves and blouses. Basic colors for motifs were arranged to create a colorful and modern but staid image in pink, blue, purple, green, yellow and brown tones. Based on a developed motif, changes were made in blouse design with lattice patterns through a variety of effects such as repetition, rotation, cross-arrangement, and oblique arrangement, and three-dimensional simulation was used to bring the design to life. Scarf design employed and applied the existing motifs in an appropriate manner for design purposes and reconstructed them through such effects as repetition, rotation, compositional variation and gradation to express a gorgeous and refined image.