• Journal of Broadcast Engineering
• /
• v.10 no.3
• /
• pp.321-327
• /
• 2005

A MVC (Multi-view Video Coding) method, which uses both an illumination change-adaptive ME (Motion Estimation)/DC (Motion Compensation) and a 2D (Dimensional) direct mode, is proposed. Firstly, a new SAD (Sum of Absolute Difference) measure for ME/MC is proposed to compensate the Luma pixel value changes for spatio-temporal motion vector prediction. Illumination change-adaptive (ICA) ME/MC uses the new SAD to improve both MV (Motion Vector) accuracy and bit saving. Secondly, The proposed 2D direct mode that can be used in inter-view prediction is an extended version of the temporal direct mode in MPEG-4 AVC. The proposed MVC method obtains approximately 0.8dB PSNR (Peak Signal-to-Noise Ratio) increment compared with the MPEG-4 AVC simulcast coding.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.12 no.6
• /
• pp.629-635
• /
• 2019

In Deep Learning models derivative is implemented by error back-propagation which enables the model to learn the error and update parameters. It can find the global (or local) optimal points of parameters even in the complex models taking advantage of a huge improvement in computing power. However, deliberately generated data points can 'fool' models and degrade the performance such as prediction accuracy. Not only these adversarial examples reduce the performance but also these examples are not easily detectable with human's eyes. In this work, we propose the method to detect adversarial datasets with random noise addition. We exploit the fact that when random noise is added, prediction accuracy of non-adversarial dataset remains almost unchanged, but that of adversarial dataset changes. We set attack methods (FGSM, Saliency Map) and noise level (0-19 with max pixel value 255) as independent variables and difference of prediction accuracy when noise was added as dependent variable in a simulation experiment. We have succeeded in extracting the threshold that separates non-adversarial and adversarial dataset. We detected the adversarial dataset using this threshold.

• Asian-Australasian Journal of Animal Sciences
• /
• v.20 no.12
• /
• pp.1901-1908
• /
• 2007

The aim of this series of experiments was to examine the opportunity for application of X-ray computer tomography (CT) in cattle production. Firstly, tissue composition of M. longissimus dorsi (LD) cuts between the $11-13^{th}$ ribs (in Exp 1. between the $9-11^{th}$ ribs), was determined by CT and correlated with tissue composition of intact half carcasses prior to dissection and tissue separation. Altogether, 207 animals of different breeds and genders were used in the study. In Exp. 2 and 3, samples were taken from LD cuts, dissected and chemical composition of muscle homogenates was analysed by conventional procedures. Correlation coefficients were calculated among slaughter records, tissues in whole carcasses and tissue composition of rib samples. Results indicated that tissue composition of rib samples determined by CT closely correlated with tissue composition results by dissection of whole carcasses. The findings revealed that figures obtained by CT correlate well with the dissection results of entire carcasses (meat, bone, fat). Close three-way coefficients of correlation (r = 0.80-0.97) were calculated among rib eye area, volume of cut, pixel-sum of adipose tissue determined by CT and intramuscular fat or adipose tissue in entire carcasses. Estimation of tissue composition of carcasses using equations including only CT-data as independent variables proved to be less reliable in prediction of lean meat and bone in carcass ($R^2 = 0.51-0.86$) than for fat (($R^2 = 0.83-0.89$). However, when cold half carcass weight was also included in the equation, the coefficient of determination exceeded $R^2 = 0.90$. In Exp. 3 tissue composition of rib samples by CT were compared to the results of EUROP carcass classification. Findings revealed that CT analysis has higher predictive value in estimation of actual tissue composition of cattle carcasses than EUROP carcass classification.

• Journal of the Institute of Convergence Signal Processing
• /
• v.10 no.1
• /
• pp.32-39
• /
• 2009

In this paper, the method of region segmentation using genetic algorithm is proposed for an improvement of efficiency in moving picture coding. A genetic algorithm is the method that searches a large probing space using only a function value for a optimal combination consecutively. By progressing both motion presumption and region segmentation at once, we can assign the motion vector in a image to a small block or a pixel respectively, and transform the capacity of coding and a signal to noise rate into a problem of optimization. That is to say, there is close correlation between region segmentation and motion presumption in motion-compensated prediction coding. This is to optimize the capacity of coding and a S/N ratio. This is to arrange the motion vector in each block of picture according to the state of optimization. Therefore, we examined both the data type of genetic algorithm and the method of data processing to obtain the results of optimal region segmentation in this paper. And we confirmed the validity of a proposed method using the test pictures by means of computer simulation.

• Smart Media Journal
• /
• v.10 no.1
• /
• pp.32-38
• /
• 2021

Image classification through deep learning on the image from photographs has been a very active research field for the past several years. In this paper, we propose a method of automatically discriminating stone images from domestic source through deep learning, which is to use Python's hash library to scan 300×300 pixel photo images of granites such as Hwangdeungseok, Goheungseok, and Pocheonseok, performing data preprocessing to create learning images by examining duplicate images for each stone, removing duplicate images with the same hash value as a result of the inspection, and deep learning by stone. In addition, to utilize VGGNet, the size of the images for each stone is resized to 224×224 pixels, learned in VGG16 where the ratio of training and verification data for learning is 80% versus 20%. After training of deep learning, the loss function graph and the accuracy graph were generated, and the prediction results of the deep learning model were output for the three kinds of stone images.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.47 no.6
• /
• pp.127-136
• /
• 2010

In multi-view video, illumination disharmony between neighboring views can occur on account of different location of each camera and imperfect camera calibration, and so on. Such discrepancy can be the cause of the performance decrease of multi-view video coding by mismatch of inter-view prediction which refer to the pictures obtained from the neighboring views at the same time. In this paper, we propose an efficient histogram-based prefiltering algorithm to compensate mismatches between the luminance and chrominance components in multi-view video for improving its coding efficiency. To compensate illumination variation efficiently, all camera frames of a multi-view sequence are adjusted to a predefined reference through the histogram matching. A Cosited filter that is used for chroma subsampling in many video encoding schemes is applied to each color component prior to histogram matching to improve its performance. The histogram matching is carried out in the RGB color space after color space converting from YCbCr color space. The effective color conversion skill that has respect to direction of edge and range of pixel value in an image is employed in the process. Experimental results show that the compression ratio for the proposed algorithm is improved comparing with other methods.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.25 no.3
• /
• pp.226-235
• /
• 2023

This study aimed to supplement the shortcomings of the Multiple-sensor-based Frost Observation System (MFOS). The developed frost observation system is an improvement of the existing system. Based on the leaf wetness sensor (LWS), it not only detects frost but also functions to predict surface temperature, which is a major factor in frost occurrence. With the existing observation system, 1) it is difficult to observe ice (frost) formation on the surface when capturing an image of the LWS with an RGB camera because the surface of the sensor reflects most visible light, 2) images captured using the RGB camera before and after sunrise are dark, and 3) the thermal infrared camera only shows the relative high and low temperature. To identify the ice (frost) generated on the surface of the LWS, a LWS that was painted black and three sheets of glass at the same height to be used as an auxiliary tool to check the occurrence of ice (frost) were installed. For RGB camera shooting before and after sunrise, synchronous LED lighting was installed so the power turns on/off according to the camera shooting time. The existing thermal infrared camera, which could only assess the relative temperature (high or low), was improved to extract the temperature value per pixel, and a comparison with the surface temperature sensor installed by the National Institute of Meteorological Sciences (NIMS) was performed to verify its accuracy. As a result of installing and operating the MFOS v2, which reflects these improvements, the accuracy and efficiency of automatic frost observation were demonstrated to be improved, and the usefulness of the data as input data for the frost prediction model was enhanced.