• The KIPS Transactions:PartB
• /
• v.11B no.2
• /
• pp.177-186
• /
• 2004

It is a weak point of the motion estimation technique for video compression that the predicted video encoding algorithm requires higher-order computational complexity. To reduce the computational complexity of encoding algorithms, researchers introduced techniques such as 3D-WT that don't require motion prediction. One of the weakest points of previous 3D-WT studies is that they require too much memory for encoding and too long delay for decoding. In this paper, we propose a technique called `FS (Fast playable and Scalable) 3D-WT' This technique uses a modified Haar wavelet transform algorithm and employs improved encoding algorithm for lower memory and shorter delay requirement. We have executed some tests to compare performance of FS 3D-WT and 3D-V. FS 3D-WT has exhibited the same high compression rate and the same short processing delay as 3D-V has.

• The Journal of the Korea Contents Association
• /
• v.12 no.8
• /
• pp.20-28
• /
• 2012

DVC(Distributed Video Coding) technique is a new paradigm, which is based on the Slepian-Wolf and Wyner-Ziv theorems. DVC offers not only flexible partitioning of the complexity between the encoder and decoder, but also robustness to channel errors due to intrinsic joint source-channel coding. Many conventional research works have been focused on the light video encoder and its rate-distortion performance improvement. However, in this paper, we propose a new DVC codec which is effectively applicable for error-prone environment. The proposed method adopts a quantiser without dead-zone and symmetric Gray code around zero value. Through computer simulations, the proposed method is evaluated by the bit errors position as well as the number of burst bit errors. Additionally, it is shown that the maximum and minimum transmission rate for the given application can be linearly determined by the number of bit errors.

• Journal of Broadcast Engineering
• /
• v.15 no.1
• /
• pp.52-62
• /
• 2010

Recently, DVC (Distributed Video Coding) is attracting a lot of research works since this enables us to implement a light-weight video encoder by distributing the high complex tasks such as motion estimation into the decoder side. In order to improve the coding efficiency of the DVC, the existing works have been focused on the efficient generation of side information (SI) or the virtual channel modeling which can describe the statistical channel noise well. But, in order to improve the overall performance, this paper proposes a new scheme that can be implemented with simple bit operations without introducing complex operation. That is, the performance of the proposed scheme is enhanced by using the fact that the Wyner-Ziv (WZ) frame and side information are highly correlated, and by reducing the effect of virtual channel noise which tends to be clustered in some regions. For this aim, this paper proposes an efficient pixel-domain WZ (PDWZ) CODEC which effectively exploits the statistical redundancy by using the code conversion and Gray code, and then reduces the channel noise by using the bit interleaver. Through computer simulations, it is shown that the proposed scheme can improve the performance up to 0.5 dB in objective visual quality.

• Journal of Broadcast Engineering
• /
• v.26 no.6
• /
• pp.778-789
• /
• 2021

Recently, various lightweight methods for using Convolutional Neural Network(CNN) models in mobile devices have emerged. Weight quantization, which lowers bit precision of weights, is a lightweight method that enables a model to be used through integer calculation in a mobile environment where GPU acceleration is unable. Weight quantization has already been used in various models as a lightweight method to reduce computational complexity and model size with a small loss of accuracy. Considering the size of memory and computing speed as well as the storage size of the device and the limited network environment, this paper proposes a method of compressing integer weights after quantization using a video codec as a method. To verify the performance of the proposed method, experiments were conducted on VGG16, Resnet50, and Resnet18 models trained with ImageNet and Places365 datasets. As a result, loss of accuracy less than 2% and high compression efficiency were achieved in various models. In addition, as a result of comparison with similar compression methods, it was verified that the compression efficiency was more than doubled.