• Journal of Broadcast Engineering
• /
• v.27 no.3
• /
• pp.283-294
• /
• 2022

The feature map used in the network for deep learning generally has larger data than the image and a higher compression rate than the image compression rate is required to transmit the feature map. This paper proposes a method for transmitting a pyramid feature map with high compression rate, which is used in a network with an FPN structure that has robustness to object size in deep learning-based image processing. In order to efficiently compress the pyramid feature map, this paper proposes a structure that predicts a pyramid feature map of a level that is not transmitted with pyramid feature map of some levels that transmitted through the proposed prediction network to efficiently compress the pyramid feature map and restores compression damage through the proposed reconstruction network. Suggested mAP, the performance of object detection for the COCO data set 2017 Train images of the proposed method, showed a performance improvement of 31.25% in BD-rate compared to the result of compressing the feature map through VTM12.0 in the rate-precision graph, and compared to the method of performing compression through PCA and DeepCABAC, the BD-rate improved by 57.79%.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2021.06a
• /
• pp.7-9
• /
• 2021

본 논문은 Detectron2 [1]에서 지원하는 객체 검출 임무 수행 네트워크의 과정 중에서 추출한 피쳐 맵을 신경망 기반으로 압축하는 방법을 제안한다. 이를 위해, 신경 망 기반 영상 압축을 지원하는 공개 소프트웨어인 CompressAI [2] 모델 중 하나인 bmshj2018-hyperprior 의 압축 네트워크를 활용하여 임무 수행 네트워크의 과정 중 스탬 레이어(stem layer)에서 추출된 피쳐 맵을 압축하도록 학습시켰다. 또한, 압축 네트워크의 입력 피쳐 맵의 너비와 높이 크기가 64 의 배수가 되도록 객체 검출 네트워크의 입력 영상 보간 값을 조정하는 방법도 제안한다. 제안하는 신경망 기반 피쳐 맵 압축 방법은 피쳐 맵을 최근 표준이 완료된 차세대 압축 표준 방법인 VVC(Versatile Video Coding, [3])로 압축한 결과에 비해 큰 성능 향상을 보이고, VCM 앵커와 유사한 성능을 보인다.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.33B no.4
• /
• pp.63-71
• /
• 1996

In this paper, a variable length block code for real time compression and decompression of hanguel and chinese character bit map font is proposed. The proposed code shows a good compression ratio in complete form of hangeul myoungjo and godik style and chinese batang and doddum style bit map font. Besides, a compression and decompression ASIC is designed and simulated on CAD. The 0.8 micron CMOS sea of gate is used to implement the ASIC in amount of 5,200 gates, and it runs at simple hardware and compress and decompress at 33M bit/sec at maximum, which is ideal for real time applications.

• Journal of Korea Multimedia Society
• /
• v.16 no.4
• /
• pp.418-429
• /
• 2013

This paper presents a GIS vector map data compression scheme based on hybrid polyline simplification method and SEC(spatial energy compaction). The proposed method extracts all layers which contain polylines in the GIS vector map and compress all polylines in extracted layers by the hybrid polyline simplification and SEC based on MAE(minimum area error) for each segment in the line. The proposed simplification and SEC increase the compression ratio while preserving the shape quality. We analyze the visual aspects and compression efficiency between the original GIS vector map and the compressed map. From experimental results, we verify that our method has the higher compression efficiency and visual quality than conventional methods.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.49 no.3
• /
• pp.30-36
• /
• 2012

block truncation coding(BTC) image compression is known as a simple and efficient technology for image compression algorithm. In this paper, we propose RMC-BTC algorithm(RMC : reduction method chrominace data) for color image compression. To compress chrominace data, in every BTC block, the RMC-BTC coding employs chrominace data expressed with average of chrominace data and using method of luminance data bit-map to represented chrominance data bit-map. Experimental results shows efficiency of proposed algorithm, as compared with PSNR and compression ratio of the conventional BTC method.

• Journal of Korea Game Society
• /
• v.9 no.6
• /
• pp.171-178
• /
• 2009

Multiple layer displacement mapping methods are able to represent more complex and general geometries which cannot be presented by single layer displacement mapping methods, and provide a realistic scene to digital contents such as 3D games and movies with relatively low costs. However, as we use more layers for details, data space is wasted more because lower layers have less displacement data than higher layers. In this paper, we suggest a lossless compression and rendering method of a multiple layer displacement map. Since we compress the map without data loss, the proposed method provides the same quality as the rendering result that uses an original multiple layer displacement map.

• Journal of Korea Multimedia Society
• /
• v.14 no.2
• /
• pp.210-218
• /
• 2011

Generally, GIS digital map has been represented and transmitted by ASCII and Binary data forms. Among these forms, Binary form has been widely used in many GIS application fields for the transmission of mass map data. In this paper, we present a hierarchical compression technique of polyline and polygon components for effective storage and transmission of vector map with various degree of decision. These components are core geometric components that represent main layers in vector map. The proposed technique performs firstly the energy compaction of all polyline and polygon components in spatial domain for the lossless compression of detailed vector map and compress independently integer parts and fraction parts of 64bit floating points. From experimental results, we confirmed that the proposed technique has superior compressive performance to the conventional data compression of 7z, zip, rar and gz.

• Journal of the Korea Convergence Society
• /
• v.9 no.10
• /
• pp.13-19
• /
• 2018

MapReduce, the Hadoop's essential core technology, is most commonly used to process big data based on the Hadoop distributed file system. However, the existing MapReduce-based big data processing techniques have a feature of dividing and storing files in blocks predefined in the Hadoop distributed file system, thus wasting huge infrastructure resources. Therefore, in this paper, we propose an efficient MapReduce-based big data processing scheme. The proposed method enhances the storage efficiency of a big data infrastructure environment by converting and compressing the data to be processed into a data format in advance suitable for processing by MapReduce. In addition, the proposed method solves the problem of the data processing time delay arising from when implementing with focus on the storage efficiency.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.39A no.7
• /
• pp.404-412
• /
• 2014

Exploiting multiple antennas at mobile devices is difficult due to limited size and power. In this paper, a distributed MIMO protocol achieving the capacity of conventinal MIMO systems is proposed and analyzed. For exploiting distributed MIMO features, Quantize-Map-and-Forward (QMF) scheme shows improved performance than Amplify-and-Forward (AF) scheme. Also, the protocol based on multiple access channel (MAC) is proposed to improve the multiplexing gain. We showed that sufficient condition of the number of slave nodes to achieve the gain of a MAC based protocol. Because the base station can support multiple clusters operating in distributed MIMO, the total cellular capacity can be extremely enhanced in proportional to the number of clusters.

• Journal of Advanced Navigation Technology
• /
• v.18 no.4
• /
• pp.401-408
• /
• 2014

Hadoop is composed with MapReduce programming model for distributed processing and HDFS distributed file system. Hadoop is suitable framework for big data processing, but processing of mass small files have many problems. The processing of mass small file in hadoop have problems to created one mapper per one file, and it have problems to needed many memory for store of meta information of file. This paper have comparison evaluation processing method of mass small file with various method in hadoop platform. The processing of general compression format is inadequate because of processing by one mapper regardless of data size. The processing of sequence and hadoop archive file is removed memory problem of namenode by compress and combine of small file. Hadoop archive file is faster then sequence file about combine time of small file. The processing using CombineFileInputFormat class is needed not combine of small file, and it have similar speed big data processing method.