• Journal of Information Processing Systems
• /
• v.17 no.4
• /
• pp.737-753
• /
• 2021

As the demand for high-quality rendering for mixed reality, videogame, and simulation has increased, global illumination has been actively researched. Monte Carlo path tracing can realize global illumination and produce photorealistic scenes that include critical effects such as color bleeding, caustics, multiple light, and shadows. If the sampling rate is insufficient, however, the rendered results have a large amount of noise. The most successful approach to eliminating or reducing Monte Carlo noise uses a feature-based filter. It exploits the scene characteristics such as a position within a world coordinate and a shading normal. In general, the techniques are based on the denoised pixel or sample and are computationally expensive. However, the main challenge for all of them is to find the appropriate weights for every feature while preserving the details of the scene. In this paper, we compare the recent algorithms for removing Monte Carlo noise in terms of their performance and quality. We also describe their advantages and disadvantages. As far as we know, this study is the first in the world to compare the artificial intelligence-based denoising methods for Monte Carlo rendering.

• 제어로봇시스템학회:학술대회논문집
• /
• 1991.10a
• /
• pp.768-772
• /
• 1991

The Self Creating and organizing (SCO) is a new architecture and one of the unsupervized learning algorithm for the artificial neural network. SCO begins with only one output node which has a sufficiently wide response range, and the response ranges of all the nodes decrease with time. Self Creating and Organizing Neural Network (SCONN) decides automatically whether adapting the weights of existing node or creating a new node. It is compared to the Kohonen's Self Organizing Feature Map (SOFM). The results show that SCONN has lots of advantages over other competitive learning architecture.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2000.11a
• /
• pp.376-379
• /
• 2000

In this paper, this system makes use of the analog infrered rays sensor and converts the feature of fish analog signal when sensor is operating with CPU(80C196KC). Then, After signal processing, this feature is classified a special feature and a outline of fish by using the neural network, one of the artificial intelligence scheme. This neural network classifies fish pattern of very simple and short calculation. This has linear activation function and the error backpropagation is used as a learning algorithm. And the neural network is learned in off-line process. Because an adaptation period of neural network is too long time when random initial weights are used, off-line learning is induced to decrease the progress time. We confirmed this method has better performance than somewhat outdated machines.

• Journal of IKEEE
• /
• v.23 no.4
• /
• pp.1166-1174
• /
• 2019

This paper propose and experiment advanced PersonNet, a human identification model, with advanced performance. We apply the inception layer to extract feature points, and increase the existing 32 feature points to 154. Also, we modify the CND method used by PersonNet to mitigate asymmetry, and apply weights to the feature map of pedestrian images in three parts, thereby making the features more distinct. Three databases were used for performance evaluation : CUHK01, CUHK03 and Market-1501. The experiment results showed 27-31% improvement in performance.

• Proceedings of the IEEK Conference
• /
• 2002.07b
• /
• pp.1236-1239
• /
• 2002

When we see an object, we usually can say what it is easily even for the case where the object isn't shown in the frontal view. However, it is difficult to believe that all views of every object we have ever seen are fully memorized in our brain. Possibly, when an object is shown, we have some typical views of the object in our brain through our past experience and reconstruct the view to recognize what the presented object is. Non-negative Matrix Factorization (NMF) is one of the methods to extract the basis images from sample data set. The prominent feature of this method is that the reconstructed image is obtained by only additions of the basis images with suitable positive weights. So NMF can be seen more biologically plausible method than any other feature extraction methods such as Vector Quantization (VQ) and principal Component Analysis (PCA). In this paper, we adopt NMF to extract the aspect features from the set of images, which consists of various views of a given object. Some experiments are shown how much well NMF can extract the aspect features than any other methods such as VQ and PCA.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.7 no.8
• /
• pp.1843-1859
• /
• 2013

The continuous increase of the spatial resolution of remote sensing images brings great challenge to image analysis and processing. Traditional prior knowledge-based region detection and target recognition algorithms for processing high resolution remote sensing images generally employ a global searching solution, which results in prohibitive computational complexity. In this paper, a more efficient region of interest (ROI) detection algorithm based on visual attention and threshold segmentation (VA-TS) is proposed, wherein a visual attention mechanism is used to eliminate image segmentation and feature detection to the entire image. The input image is subsampled to decrease the amount of data and the discrete moment transform (DMT) feature is extracted to provide a finer description of the edges. The feature maps are combined with weights according to the amount of the "strong points" and the "salient points". A threshold segmentation strategy is employed to obtain more accurate region of interest shape information with the very low computational complexity. Experimental statistics have shown that the proposed algorithm is computational efficient and provide more visually accurate detection results. The calculation time is only about 0.7% of the traditional Itti's model.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.8 no.6
• /
• pp.1135-1141
• /
• 2004

In this study, this system makes use of the analog infrared rays sensor and converts the feature of fish outline when sensor is operating with CPU(80C196KC). Then, after signal processing, this feature is classified a special feature and a outline of fish by using the neural network, one of the artificial intelligence scheme. This neural network classifies fish pattern of very simple and short calculation. This has linear activation function and the error back propagation is used as a teaming algorithm. And the neural network is learned in off-line process. Because an adaptation period of neural network is too long when random initial weights are used, off-line teaming is induced to decrease the progress time.

• The KIPS Transactions:PartB
• /
• v.15B no.5
• /
• pp.457-464
• /
• 2008

In the real-world operational environment, most of text classification systems have the problems of insufficient training documents and no prior knowledge of feature space. In this regard, $Na{\ddot{i}ve$ Bayes is known to be an appropriate algorithm of operational text classification since the classification model can be evolved easily by incrementally updating its pre-learned classification model and feature space. This paper proposes the improving technique of $Na{\ddot{i}ve$ Bayes classifier through feature weighting strategy. The basic idea is that parameter estimation of $Na{\ddot{i}ve$ Bayes considers the degree of feature importance as well as feature distribution. We can develop a more accurate classification model by incorporating feature weights into Naive Bayes learning algorithm, not performing a learning process with a reduced feature set. In addition, we have extended a conventional feature update algorithm for incremental feature weighting in a dynamic operational environment. To evaluate the proposed method, we perform the experiments using the various document collections, and show that the traditional $Na{\ddot{i}ve$ Bayes classifier can be significantly improved by the proposed technique.

• Journal of Korea Multimedia Society
• /
• v.14 no.6
• /
• pp.742-751
• /
• 2011

This paper presents a robust 3D model hashing dependent on key and parameter by using heat kernel signature (HKS), which is special shape feature descriptor, In the proposed hashing, we calculate HKS coefficients of local and global time scales from eigenvalue and eigenvector of Mesh Laplace operator and cluster pairs of HKS coefficients to 2D square cells and calculate feature coefficients by the distance weights of pairs of HKS coefficients on each cell. Then we generate the binary hash through binarizing the intermediate hash that is the combination of the feature coefficients and the random coefficients. In our experiment, we evaluated the robustness against geometrical and topological attacks and the uniqueness of key and model and also evaluated the model space by estimating the attack intensity that can authenticate 3D model. Experimental results verified that the proposed scheme has more the improved performance than the conventional hashing on the robustness, uniqueness, model space.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.39 no.3
• /
• pp.32-43
• /
• 2002

In this paper, we propose a pattern classification system for digital signal which is based on neural networks. The proposed system consists of two models of neural network. The first part is a wavelet neural network whose role is a feature extraction. For this part, we compare existing models of wavelet networks and propose a new model for feature extraction. The other part is a wavelet network for pattern classification. We modify the structure of previous wavelet network for pattern classification and propose a learning method. The inputs of the pattern classification wavelet network is connection weights, dilation and translation parameters in hidden nodes of feature extraction network. And the output is a class of the signal which is input of feature extraction network. The proposed system is, applied to classification of EEG signal based on frequency.