- 學位論文
基於GPU平行DPCM視訊壓縮編碼法
Parallel DPCM Video Compression Based on GPU
摘要
隨著網路的快速發展，許多有線系統也漸漸發展出無線的系統，如無線汽車後視系統，無線監視系統，視訊電話等，對於尚未安裝設備的人而言，採用無線系統會更方便於安裝。 在無線視頻系統上必須考量，(1)每秒顯示幀數(frames per second，FPS)(2)影像品質；若使用複雜性較高的壓縮編碼技術，會讓系統的運算時間增加，以至於無法達到即時反應，因此考量重點會在兩者之間的平衡。 隨著硬體架構與繪圖處理器(Graphics Processing Unit，GPU)的快速發展，基於GPU高速運算技術出現，讓視頻能同時滿足高畫面更新率與影像品質，提供了一條可行的途徑。 本論文利用GPU技術完成誤差脈衝編碼調變(Differential Pulse Code Modulation，DPCM)與影像處理的平行化，將編解碼與影像縮放的步驟移植到GPU上執行，以實現GPU平行計算。 經研究證明和分析，利用GPU架構最佳化的DPCM影像視訊壓縮編碼器，在執行時間上具有更快的速度，比起只有使用CPU做運算皆有顯著的效能提升，並且可以提高每秒顯示幀數，視頻也可以維持在一定的品質。
並列摘要
With the rapid development of Internet, Many cable systems also gradually developed a wireless system, Such as wireless car rearview system, Wireless Monitoring System, Video phone, etc. For yet device, The use of wireless systems will be more convenient to installm. Must be considered in the video reversing system: FPS and Image quality; If you are using the complexity of higher compression technology, will leave the system computing time increases, that they can not achieve the immediate reaction, therefore consider the focus will balance between the two. With the rapid development of the hardware architecture and GPU, base GPU high-speed computing technology appears, Video can also meet the high frame rates and image quality, Provides a feasible way. In this thesis, using GPU technology to complete DPCM and Image, parallel to the image processing, Code decoding and image scaling steps ported to the GPU execution, in order to achieve the GPU parallel computing. The research shows and analysis, use GPU architecture optimized DPCM, have faster speed in the execution time, both a significant performance increase compared to only using the CPU to do calculations, and can improve FPS, the video can also be maintained at a certain quality