• Journal of Korea Multimedia Society
• /
• v.22 no.12
• /
• pp.1457-1465
• /
• 2019

3D convolution is to stack multiple consecutive frames to form a cube, and then apply the 3D convolution kernel in the cube. In this structure, each feature map of the convolutional layer is connected to multiple adjacent sequential frames in the previous layer, thus capturing the motion information. However, due to the changes of pedestrian posture, motion and position, the convolution at the same place is inappropriate, and when the 3D convolution kernel is convoluted in the time domain, only time domain features of three consecutive frames can be extracted, which is not a good enough to get action information. This paper proposes an action recognition method based on feature fusion of 3D convolutional neural network. Based on the VGG16 network model, sending a pre-acquired optical flow image for learning, then get the time domain features, and then the feature of the time domain is extracted from the features extracted by the 3D convolutional neural network. Finally, the behavior classification is done by the SVM classifier.

• Journal of Broadcast Engineering
• /
• v.23 no.5
• /
• pp.669-681
• /
• 2018

CNN (Convolution Neural Network) is one of the most important techniques to identify the kind of objects in the captured pictures. Whereas the conventional models have been used for low resolution images, the technique to recognize the high resolution images becomes crucial in the field of artificial intelligence. In this paper, we proposed an efficient CNN model based on dilated convolution and thresholding techniques to increase the recognition ratio and to decrease the computational complexity. The simulation results show that the proposed algorithm outperforms the conventional method and the thresholding technique enhances the performance of the proposed model.

• Smart Structures and Systems
• /
• v.17 no.5
• /
• pp.753-771
• /
• 2016

This paper addresses the free and transient responses of a SDOF linear complex stiffness system by making use of the Hilbert transform and the convolution integral. Because the second-order differential equation of motion having the complex stiffness give rise to the conjugate complex eigen values, its time-domain analysis using the standard time integration scheme suffers from the numerical instability and divergence. In order to overcome this problem, the transient response of the linear complex stiffness system is obtained by the convolution integral of a green function which corresponds to the unit-impulse free vibration response of the complex system. The damped free vibration of the complex system is theoretically derived by making use of the state-space formulation and the Hilbert transform. The convolution integral is implemented by piecewise-linearly interpolating the external force and by superimposing the transient responses of discretized piecewise impulse forces. The numerical experiments are carried out to verify the proposed time-domain analysis method, and the correlation between the real and imaginary parts in the free and transient responses is also investigated.

• The Journal of the Acoustical Society of Korea
• /
• v.15 no.5
• /
• pp.13-20
• /
• 1996

Recently, for implementation of neural networks extensive studies have been done especially VLSI technology has been regarded as the one of the most attractive means to implement neural networks. The main drawbacks of digital VLSI implementations are their large area and slow processing speed. In this paper to solve the speed and size problems we designed the efficient architecture using the binary convolution method for basic operation of neural cell, multiplication and addition. When it is used for implementing 3-layer network with 16 neural cell per layer that used neural cell based on binary convolution, clock of 50MHz and 26MCPS on 0.8${\mu}$ standard cell library has been achieved.

• The Journal of Korean Association of Computer Education
• /
• v.11 no.5
• /
• pp.57-66
• /
• 2008

An adaptive optimization of parametric cubic convolution for image interpolation is derived in this paper. The proposed technique is based on optimizing the standard cubic convolution interpolation formula at each interpolated pixel. Conventional parametric cubic convolution methods use a fixed parameter in an image, so properties of each pixel cannot be incorporated into the interpolation. The proposed method optimizes the interpolation kernel by obtaining parameters adaptively on each pixel. A new cost function is introduced to reflect frequency properties of the original data. The proposed technique produces noticeably sharper edges than traditional techniques and exhibits an average PSNR improvement of traditional techniques.

• ETRI Journal
• /
• v.42 no.2
• /
• pp.230-238
• /
• 2020

Human-object interaction (HOI) detection is a popular computer vision task that detects interactions between humans and objects. This task can be useful in many applications that require a deeper understanding of semantic scenes. Current HOI detection networks typically consist of a feature extractor followed by detection layers comprising small filters (eg, 1 × 1 or 3 × 3). Although small filters can capture local spatial features with a few parameters, they fail to capture larger context information relevant for recognizing interactions between humans and distant objects owing to their small receptive regions. Hence, we herein propose a three-stream HOI detection network that employs a context convolution module (CCM) in each stream branch. The CCM can capture larger contexts from input feature maps by adopting combinations of large separable convolution layers and residual-based convolution layers without increasing the number of parameters by using fewer large separable filters. We evaluate our HOI detection method using two benchmark datasets, V-COCO and HICO-DET, and demonstrate its state-of-the-art performance.

• Journal of Broadcast Engineering
• /
• v.13 no.5
• /
• pp.707-718
• /
• 2008

We derive a modified version of the cubic convolution scaler to enlarge or reduce the size of digital image with arbitrary ratio. To enhance the edge information of the scaled image and to obtain a high-quality scaled image, the proposed scaler is applied along the direction of an edge. Since interpolation along the direction of an edge has to process nonuniformly sampled data, the kernel of the cubic convolution scaler is modified to interpolate the data. The proposed scaling scheme can be used to resize pictures in various formats in a transcoding system that transforms a bit stream compressed at one bit rate into one compressed at another bit rate. In many applications, such as transcoders, the resolution conversion is very important for changing the image size while maintaining high quality of the scaled image. We show experimental results that demonstrate the effectiveness of the proposed interpolation method. The proposed scheme provides clearer edges, without artifacts, in the resized image than do conventional schemes. The algorithm exhibits significant improvement in the minimization of information loss when compared with the conventional interpolation algorithms.

• International Journal of Computer Science & Network Security
• /
• v.21 no.9
• /
• pp.203-211
• /
• 2021

Image tampering detection and localization have become an active area of research in the field of digital image forensics in recent times. This is due to the widespread of malicious image tampering. This study presents a new method for image tampering detection and localization that combines the advantages of dilated convolution, residual network, and UNET Architecture. Using the UNET architecture as a backbone, we built the proposed network from two kinds of residual units, one for the encoder path and the other for the decoder path. The residual units help to speed up the training process and facilitate information propagation between the lower layers and the higher layers which are often difficult to train. To capture global image tampering artifacts and reduce the computational burden of the proposed method, we enlarge the receptive field size of the convolutional kernels by adopting dilated convolutions in the residual units used in building the proposed network. In contrast to existing deep learning methods, having a large number of layers, many network parameters, and often difficult to train, the proposed method can achieve excellent performance with a fewer number of parameters and less computational cost. To test the performance of the proposed method, we evaluate its performance in the context of four benchmark image forensics datasets. Experimental results show that the proposed method outperforms existing methods and could be potentially used to enhance image tampering detection and localization.

• Journal of Korea Multimedia Society
• /
• v.16 no.5
• /
• pp.637-646
• /
• 2013

In order to model a variety of natural trees that are appropriate to outdoor terrains consisting of multiple trees, this study proposes a modeling method of new growth rules(based on the convolution sums of divisor functions). Basically, this method uses an existing growth-volume based algorithm for efficient management of the branches and leaves that constitute a tree, as well as natural propagation of branches. The main features of this paper is to introduce the theory of convolution sums of divisor functions that is naturally expressed the growth or fate of branches and leaves at each growth step. Based on this, a method of modeling various tree is proposed to minimize user control through a number of divisor functions having generalized generation functions and modification of the growth rule. This modeling method is characterized by its consideration of both branches and leaves as well as its advantage of having a greater effect on the construction of an outdoor terrain composed of multiple trees. Natural and varied tree model creation through the proposed method was conducted, and using this, the possibility of constructing a wide nature terrain and the efficiency of the process for configuring multiple trees were evaluated experimentally.

• Journal of Software Assessment and Valuation
• /
• v.15 no.2
• /
• pp.59-66
• /
• 2019

Since Jeju is the absolute weight of agriculture and tourism, the analysis of precipitation is more important than other regions. Currently, some numerical models are used for analysis of precipitation of Jeju Island using observation data from meteorological satellites. However, since precipitation changes are more diverse than other regions, it is difficult to obtain satisfactory results using the existing numerical models. In this paper, we propose a Jeju precipitation pattern analysis method using the texture analysis method based on Convolution Neural Network (CNN). The proposed method converts the water vapor image and the temperature information of the area of ​​Jeju Island from the weather satellite into texture images. Then converted images are fed into the CNN to analyse the precipitation patterns of Jeju Island. We implement the proposed method and show the effectiveness of the proposed method through experiments.