• The KIPS Transactions:PartB
• /
• v.10B no.4
• /
• pp.397-402
• /
• 2003

This paper proposes an efficient fixed-point (FP) algorithm for improving performances of the independent component analysis (ICA) based on neural networks. The proposed algorithm is the FP algorithm based on Newton method for ICA using the adaptive learning parameters. The purpose of this algorithm is to improve the separation speed and performance by using the learning parameters in Newton method, which is based on the first order differential computation of entropy optimization function. The learning rate and the moment are adaptively adjusted according to an updating state of inverse mixing matrix. The proposed algorithm has been applied to the fingerprints and the images generated by random mixing matrix in the 8 fingerprints of 256${\times}$256-pixel and the 10 images of 512$\times$512-pixel, respectively. The simulation results show that the proposed algorithm has the separation speed and performance better than those using the conventional FP algorithm based on Newton method. Especially, the proposed algorithm gives relatively larger improvement degree as the problem size increases.

• Journal of Broadcast Engineering
• /
• v.10 no.4 s.29
• /
• pp.621-631
• /
• 2005

To enhance the performance of multi-view sequence CODEC, an efficient disparity vector coding method fur multiview sequences is proposed herein. For higher coding efficiency, we encode the differential vectors acquired by subtracting the original vectors from the predicted ones. To enhance the performance of disparity vector coding, it is essential to predict the disparity vectors accurately. The prediction by this proposed method utilizes the correlation among the multiview images, while conventional methods exploit the correlation among the causal blocks. Experiments were performed fur three different 5 view sequences. We were able to confirm that the proposed method predicts disparity vectors accurately by comparing the entropy and the mean absolute values for differential vectors with conventional methods. Its performance is superior to vector coding methods used in MPEG-4 which uses only a spatial correlation. The proposed method increases the coding efficiency by a factor of $30{\~}45\%$ while preserving image quality.

• Journal of Korea Multimedia Society
• /
• v.16 no.12
• /
• pp.1376-1384
• /
• 2013

In this paper, we propose an automaticdetection method of foreign bodies through template matching in industrial CT volume data. Our method is composed of three main steps. First,Indown-sampling data, the product region is separated from background after noise reduction and initial foreign-body candidates are extracted using mean and standard deviation of the product region. Then foreign-body candidates are extracted using K-means clustering. Second, the foreign body with different intensity of product region is detected using template matching. At this time, the template matching is performed by evaluating SSD orjoint entropy according to the size of detected foreign-body candidates. Third, to improve thedetection rate of foreign body in original volume data, final foreign bodiesare detected using percolation method. For the performance evaluation of our method, industrial CT volume data and simulation data are used. Then visual inspection and accuracy assessment are performed and processing time is measured. For accuracy assessment, density-based detection method is used as comparative method and Dice's coefficient is measured.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.32 no.3C
• /
• pp.297-305
• /
• 2007

In the paper, a Motion JPEG2000 coder which has been set as the standard for image compression by the Digital Cinema Initiatives (DCI), an organization composed of major movie studios was implemented into a target FPGA. The DWT (Discrete Wavelet Transform) based on lifting and the Tier 1 of EBCOT (Embedded Block Coding with Optimized Truncation) which are major functional modules of the JPEG2000 were setup with dedicated hardware. The Tier 2 process was implemented in software. For digital cinema the tile-size was set to support $1024\times1024$ pixels. To ensure the real-time operations, three entropy encoders were used. When Verilog-HDL was used for hardware, resources of 32,470 LEs in Altera's Stratix EP1S80 were used, and the hardware worked stably at the frequency of 150Mhz.

• Journal of the Korea Society of Computer and Information
• /
• v.19 no.4
• /
• pp.55-62
• /
• 2014

This paper proposes a fault detection method utilizing converted images of acoustic emission signals and texture analysis for identifying bearing's multi-faults which frequently occur in an induction motor. The proposed method analyzes three texture features from the converted images of multi-faults: multi-faults image's entropy, homogeneity, and energy. These extracted features are then used as inputs of a fuzzy-ARTMAP to identify each multi-fault including outer-inner, inner-roller, and outer-roller. The experimental results using ten times trials indicate that the proposed method achieves 100% accuracy in the fault classification.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.7 no.4
• /
• pp.163-172
• /
• 2012

A driver assistance system is critical to improve a convenience and stability of vehicle driving. Several systems have been already commercialized such as adaptive cruise control system and forward collision warning system. Efficient vehicle detection is very important to improve such driver assistance systems. Most existing vehicle detection systems are based on a radar system, which measures distance between a host and leading (or oncoming) vehicles under various weather conditions. However, it requires high deployment cost and complexity overload when there are many vehicles. A camera based vehicle detection technique is also good alternative method because of low cost and simple implementation. In general, night time vehicle detection is more complicated than day time vehicle detection, because it is much more difficult to distinguish the vehicle's features such as outline and color under the dim environment. This paper proposes a method to detect vehicles at night time using analysis of a captured color space with reduction of reflection and other light sources in images. Four colors spaces, namely RGB, YCbCr, normalized RGB and Ruta-RGB, are compared each other and evaluated. A suboptimal threshold value is determined by Otsu algorithm and applied to extract candidates of taillights of leading vehicles. Statistical features such as mean, variance, skewness, kurtosis, and entropy are extracted from the candidate regions and used as feature vector for SVM(Support Vector Machine) classifier. According to our simulation results, the proposed statistical feature based SVM provides relatively high performances of leading vehicle detection with various distances in variable nighttime environments.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.14 no.8
• /
• pp.3295-3311
• /
• 2020

With the wide application of hyperspectral images, it becomes more and more important to compress hyperspectral images. Conventional recursive least squares (CRLS) algorithm has great potentiality in lossless compression for hyperspectral images. The prediction accuracy of CRLS is closely related to the correlations between the reference bands and the current band, and the similarity between pixels in prediction context. According to this characteristic, we present an improved CRLS with adaptive band selection and adaptive predictor selection (CRLS-ABS-APS). Firstly, a spectral vector correlation coefficient-based k-means clustering algorithm is employed to generate clustering map. Afterwards, an adaptive band selection strategy based on inter-spectral correlation coefficient is adopted to select the reference bands for each band. Then, an adaptive predictor selection strategy based on clustering map is adopted to select the optimal CRLS predictor for each pixel. In addition, a double snake scan mode is used to further improve the similarity of prediction context, and a recursive average estimation method is used to accelerate the local average calculation. Finally, the prediction residuals are entropy encoded by arithmetic encoder. Experiments on the Airborne Visible Infrared Imaging Spectrometer (AVIRIS) 2006 data set show that the CRLS-ABS-APS achieves average bit rates of 3.28 bpp, 5.55 bpp and 2.39 bpp on the three subsets, respectively. The results indicate that the CRLS-ABS-APS effectively improves the compression effect with lower computation complexity, and outperforms to the current state-of-the-art methods.

• Journal of the Korean Society of Radiology
• /
• v.10 no.8
• /
• pp.645-652
• /
• 2016

In this study, based on the analysis of texture feature values of statistical properties. And we examined the normal and the applicability of the computer-aided diagnosis of cerebral infarction in the brain computed tomography images. The experiment was analyzed to evaluate the ROC curve recognition rate of disease using six parameters representing the feature values of the texture. As a result, it showed average mean 88%, variance 92%, relative smoothness 94%, uniformity of 88%, a high disease recognition rate of entropy 84%. However, it showed a slightly lower disease recognition rate and 58% for skewness. In the analysis using ROC curve, the area under the curve for each parameter indicates 0.886 (p = 0.0001) or more, resulted in a meaningful recognition of the disease. Further, to determine the cut-off values for each parameter are determined to be the prediction of disease through the computer-aided diagnosis.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.28 no.11C
• /
• pp.1118-1122
• /
• 2003

In this paper, we propose a new invisible digital watermarking scheme based on wavelet transform using singular value decomposition. Embedding process is started by decomposing the lowest frequency band image with 3${\times}$3 block among which we define the watermark block chosen by a key set; entropy and condition number of the block. A watermark is embedded in the singular values of each watermark blocks. This provides a robust watermarking in lowest possible time-frequency domain. To detect the watermark, we are locally modeling an attack as 3${\times}$3 matrices on the watermark blocks. Combining with the SVD and the attack matrices, we estimate watermark set corresponding to the watermark blocks. In each watermark block, we determine an optimal watermark which is justified by the T-testing. A numerical experiment shows that the proposed watermarking scheme efficiently detects the watermarks from several JPEG attacks.

• KIPS Transactions on Computer and Communication Systems
• /
• v.1 no.3
• /
• pp.187-192
• /
• 2012

This paper describes efficient H.264/AVC video scrambling methods for digital rights management. The proposed scrambling methods are to scramble level and suffix in entropy encoding and MVD in motion estimation of the H.264 video compression process. Other scrambling methods have been proposed but they degrade the compression efficiency or make it difficult to achieve real-time processing due to the large amount of computational efforts. Since the proposed scrambling methods resolve the drawbacks of other approaches, they do not cause image distortion and the original compression efficiency is maintained. We verified our scrambling methods and evaluated the performance by conducting several experiments with H.264 reference program. Finally, we implemented video player system using USB dongle in order to apply the proposed scrambling/descrambling methods to H.264 video compression.