先讀入膚色，非膚色的圖片，在將各張圖片由RGB轉成YCbCr。讀進來的時候，我是用四維的陣列去存膚色的圖片以及非膚色的圖片。第一個維度是用來存放有幾張圖片，接下來是解析度，還有幾個channel。讀進來之後再分別對膚色，非膚色的Cb，Cr來作統計。分別會得到Cb，Cr的mean和variance。再依次分別帶入Multivariate Normal Density的公式去求得膚色，非膚色出現的機率。最後再將得到的兩個機率值相除，得到的值，拿來和threshold值相比，藉此來判斷Input的圖的pixel是否為膚色。-To read into the color, non-color pictures, all pictures in the RGB conversion from YCbCr. Reading came when I was using four-dimensional array to save color images and non-color pictures. The first dimension is used to store a few pictures, followed by the resolution, there are several channel. Respectively after reading came in color, non-color of the Cb, Cr to make the statistical. Respectively, will be Cb, Cr of the mean and variance. Then in turn were brought into the Multivariate Normal Density formula to obtain color, the probability of non-skin color appears. Finally got two chances and then divided the value obtained values, brought, and compared with threshold values, thereby to determine the Input of the map whether the color pixel.