以網路攝影機為視訊來源的網路視訊監控系統(Network Video Recorder; NVR)是安全監控系統的主流,而所使用視訊壓縮標準為H.264。雖然H.264提供高壓縮率與高品質畫面,但複雜的解壓縮處理卻使得同時對多視訊串流解碼變成一項難解問題。本論文提出一個在頻率域下調降H.264影像解析度的方法,藉此來加快解碼速度。由於H.264編碼方式是針對冗餘值(residual)為處理對象,因此在降解析度的過程必須先還原預測值(predictor)得出真實值(real value)。我們利用離散餘弦轉換之特性,將原本須在像素域還原的預測值改至頻率域中還原,而降解析的方法為在頻率域中擷取低頻係數來取代在像素域中使用線性內插。另外我們也提出方法,藉由DC值的提取與還原,更進一步降低影像解析度來加快處理速度。實驗結果驗證所提方法的正確性與有效。
The main objective of our research is to present a method to increase H.264 decoding efficiency by downsizing image resolution in frequency domain. We proposed a method to recover the predictor in the frequency domain instead of the spatial domain. We extract the low-frequency part of the DCT block to make a downsized version instead of using bilinear algorithm. Besides, we proposed another method by extracting only the DC value of the image to downsize further and also increase more decoding efficiency.