• Journal of Broadcast Engineering
• /
• v.14 no.1
• /
• pp.59-69
• /
• 2009

A bi-directional symmetric prediction technique has been developed to improve coding efficiency of B-slice and to reduce the computational complexity required to estimate two motion vectors. On the contrary to the conventional bi-directional mode which encodes both forward and backward motion vectors, it only encodes a single forward motion vector, and the missing backward motion vector is derived in a symmetric way from the forward motion vector using temporal distance between forward/backward reference frames to and from the current B picture. Since the backward motion vector is derived from the forward motion vector, it can halve the computational complexity for motion estimation, and also reduces motion vector data to encode. This technique always derives the backward motion vector from the forward motion vector, however, there are cases when the forward motion vector is better to be derived from the backward motion vector especially in scene changes. In this paper, we generalize the idea of the symmetric coding with forward motion vector coding, and propose a new symmetric coding with backward motion vector coding and adaptive selection between the conventional symmetric mode and the proposed symmetric mode based on rate-distortion optimization.

• Korean Journal of Remote Sensing
• /
• v.35 no.6_1
• /
• pp.959-971
• /
• 2019

CNN (Convolutional Neural Network) is one representative deep learning algorithm, which can extract high-level spatial and spectral features, and has been applied for hyperspectral image classification. However, one significant drawback behind the application of CNNs in hyperspectral images is the high dimensionality of the data, which increases the training time and processing complexity. To address this problem, several CNN based hyperspectral image classification studies have exploited PCA (Principal Component Analysis) for dimensionality reduction. One limitation to this is that the spectral information of the original image can be lost through PCA. Although it is clear that the use of PCA affects the accuracy and the CNN training time, the impact of PCA for CNN based hyperspectral image classification has been understudied. The purpose of this study is to analyze the quantitative effect of PCA in CNN for hyperspectral image classification. The hyperspectral images were first transformed through PCA and applied into the CNN model by varying the size of the reduced dimensionality. In addition, 2D-CNN and 3D-CNN frameworks were applied to analyze the sensitivity of the PCA with respect to the convolution kernel in the model. Experimental results were evaluated based on classification accuracy, learning time, variance ratio, and training process. The size of the reduced dimensionality was the most efficient when the explained variance ratio recorded 99.7%~99.8%. Since the 3D kernel had higher classification accuracy in the original-CNN than the PCA-CNN in comparison to the 2D-CNN, the results revealed that the dimensionality reduction was relatively less effective in 3D kernel.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.19 no.6
• /
• pp.634-638
• /
• 2018

Haze is a factor that degrades the performance of various image processing algorithms, such as those for detection, tracking, and recognition using an electro-optical sensor. For robust operation of an electro-optical sensor-based unmanned system used outdoors, an algorithm capable of effectively removing haze is needed. As a haze removal method using a single electro-optical sensor, the dark channel prior using statistical properties of the electro-optical sensor is most widely known. Previous methods used a square filter in the process of obtaining a transmission using the dark channel prior. When a square filter is used, the effect of removing haze becomes smaller as the size of the filter becomes larger. When the size of the filter becomes excessively small, over-saturation occurs, and color information in the image is lost. Since the size of the filter greatly affects the performance of the algorithm, a relatively large filter is generally used, or a small filter is used so that no over-saturation occurs, depending on the image. In this paper, we propose an improved haze removal method using color image segmentation. The parameters of the color image segmentation are automatically set according to the information complexity of the image, and the over-saturation phenomenon does not occur by estimating the amount of transmission based on the parameters.

• Korean Journal of Remote Sensing
• /
• v.20 no.2
• /
• pp.91-102
• /
• 2004

This paper reflects the estimation of using the EOC(Electro-optical Camera) images supporting GIUH(geomorphological instantaneous unit hydrograph) approach. We have analyzed GIUH in its density and frequency distribution by creating a DEM(digital elevation model) for the sub basin produced from the EOC images and examined topographical and hydrological application possibility of the EOC images. In this process, we have topographical basin characteristic analysis that use the remote sensing technique analyzing the DEM creation process of the EOC stereo images by studying the basic topographical hydrology analysis about abstraction technique since it is flirty complex and is more time-consuming than other method. we executed statistical analysis of a basin size and river length using the frequency function after divided lattice spacing applied have to the sub river basin from the image data and the digital map into 10m intervals ranging from 10m to 100m. After comparing and examining the peak and time to peak of the GIUH, we proceeded with a comparative analysis by lattice concerning the topographical divergence rate, area ratio, length ratio. Accumulating the peak and time to peak of the GIUH is altered to non-linear form in accordance to lattice dimension as well as basin factor. It was proved that the lattice dimension is one of the important factors about the peak and time to peak of the GIUH.

• Journal of Broadcast Engineering
• /
• v.19 no.5
• /
• pp.713-725
• /
• 2014

Multi-view and 3D video technologies for a next generation video service are widely studied. These technologies can make users feel realistic experience as supporting various views. Because acquisition and transmission of a large number of views require a high cost, main challenges for multi-view and 3D video include view synthesis, video coding, and depth coding. Recently, JCT-3V (joint collaborative team on 3D video coding extension development) has being developed a new standard for multi-view and 3D video. In this paper, major tools adopted in this standard are introduced and evaluated in terms of coding efficiency and complexity. This performance analysis would be helpful for the development of a fast 3D video encoder as well as a new 3D video coding algorithm.

• Journal of Broadcast Engineering
• /
• v.2 no.2
• /
• pp.86-93
• /
• 1997

Fast growth of multimedia applications requires new functions for video data processing. such as obj;cted-based video representation and manipulation. which are not supported by 11PEG-l and 11PEG-2. To support these requirements. 11PEG-4 video coding allows users to manipulate every video object easily by decomposing a scene into several video objects and coding each of them independently. However. the large amount of computations and flexible structure of 11PEG-4 video CODEC make it difficult to be implemented by either the general purpose DSP or a dedicated VLSI. In this paper, we propose a hardware architecture using a hybrid of a high performance programmable DSP and an application specific IC to implement a flexible 11PEG-4 video codec requiring the large amount of computations. The application specific IC has the functions of motion estimation and compensation.

• Journal of the Korea Computer Graphics Society
• /
• v.22 no.3
• /
• pp.89-99
• /
• 2016

One of the fundamental elements in developing mixed reality applications is to effectively analyze and apply the environmental lighting information to image synthesis. In particular, interactive applications require to process dynamically varying lighting sources in real-time, reflecting them properly in rendering results. Previous related works are not often appropriate for this because they are usually designed to synthesize photorealistic images, generating too many, often exponentially increasing, light sources or having too heavy a computational complexity. In this paper, we present a fast light source estimation technique that aims to search for primary light sources on the fly from a sequence of video images taken by a camera equipped with a fisheye lens. In contrast to previous methods, our technique can adust the number of found light sources approximately to the size that a user specifies. Thus, it can be effectively used in Phong-illumination-model-based direct illumination or soft shadow generation through light sampling over area lights.

• Spatial Information Research
• /
• v.19 no.4
• /
• pp.73-79
• /
• 2011

With the recent increase in demand for geo-registered imagery, Mobile Mapping Systems(MMS), which can quickly construct geographic information, has become important. The main part of MMS is the high-precision observation system, which collects geographic information at a certain speed. MMS has a complex data generation process and requires a standard-specific vehicle for its use, limiting its application range. In this paper, lightweight MMS is proposed to overcome its complexity by replacing the time synchronizer with a high-speed camera and by stabilizing motion with MEMS IMU/GPS. The proposed low-cost, portable method is expected to produce of geo-registered imagery efficiently.

• Journal of the Korean Association of Geographic Information Studies
• /
• v.14 no.4
• /
• pp.13-21
• /
• 2011

Update of road facilities database such as road sign, traffic lights, and street lights is interesting business in a local government. Recently, existing road facilities database, aerial photo and topographic map are referred for the installation and complement of road facilities. But it is difficult to comprehend road facilities' condition and additional expenses may appear in field survey. Therefore, it is necessary to establish and update road facility DB and many studies has been carried out to efficiently collect road related spatial data. In this study, the establishment of various complicated road facility DB was conducted by images that had been obtained by digital camera with a built-in bluetooth and DGPS. Results showed that road facility DB was constructed effectively and suggested the possibility of road facility management using images based on coordinate through accuracy analyses using total-station surveying. And using digital camera and DGPS is expected to effective real-time update and management of road facility DB.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.9 no.10
• /
• pp.1207-1214
• /
• 2014

The first frame of a GOP is encoded in intra mode which generates a larger number of bits. In addition, the first frame is used for the inter mode encoding of the following frames. Thus the encoding results of the intra frame affects the first frame as well as the following frames. Traditionally, the quantization parameter for an intra frame is determined only depending on the bpp not considering the characteristics of the intra frame. For accurate intra frame encoding, we should consider not only bpp but also the complexity of the video sequence and the output bandwidth. In this paper, we propose a real-time quantization model which is used to calculate the quantization parameter for an intra frame encoding based on the investigation on the characteristics of a GOP. It is shown by experimental results that the proposed quantization model captures the characteristics of an intra frame effectively and the proposed method for model parameters accurately estimates the real values.