移動偵測（Motion Estimation; ME）是視訊編碼中相當重要的一種編碼工具。它利用了影像與影像之間的相似性，在連續影像間的移動向量差異不大的情況下，藉由精確的移動向量預測值產生與原始影像最相似的重組影像，再將重組影像與原始影像取差值得到最小的殘值以達到最佳的壓縮效果。本文提出一種利用中位數電路搜尋空間移動向量之中位數，作為選擇移動偵測起始位置的方法。依據邊界條件的判斷方法，決定目前區塊合適的預測模式。不同的預測模式影響（1）選取與目前區塊相鄰巨方塊不同區塊分割模式移動向量作為參考向量的方法（2）中位數電路的核心架構。以Median-of-4為核心的中位數電路可延伸為Median-of-6中位數電路。由實驗數據顯示，我們所提出的方法在八段不同解析度的影片中，於低、高量化參數設定下分別可得到0.4%的及2 %的壓縮率改善。

Motion Estimation (ME) is an important encoding tool for video compression. With the correlation and similar Motion-vectors between adjacent images, ME obtain the most similar reconstruct image and realizes the best compression through the minimum residue. An initial search point selection method for motion estimation is disclosed. The initial search point is set by the median obtained from the plurality of spatial motion vectors. The adequate prediction mode for current block is based on its boundary condition. Both the select method of best motion vector candidates for current block, and the main component of the plurality are affected by prediction mode. The Median-of-6 plurality is extended from Median-of-4 plurality. According to the experiment results, the bitstream of eight test sequences with different resolution can get 0.4 percentage and 2 percentage improvement under low-QP and high-QP environment.