alpha混合的数学原理与实现

玩davinci的人都知道，视频处理后端（视频图像显示）中的attribute层，是VIDEO1和OSD0的alpha混合，是framebuffer的驱动实现的这种效果。

    alpha混合技术，主要是用于实现半透明的效果， 假设一种不透明东西的颜色是A（即VIDEO1），另一种透明的东西的颜色是B（即OSD0），那么透过B去看A，看上去的颜色C（即attribute）就是B和A的混合颜色，可以用下面的式子来近似，设B物体的透明度为alpha(取值为0~1之间的浮点数，0为完全透明，1为完全不透明)^[1]。

    R(C)=alpha*R(B)+(1-alpha)*R(A)
    G(C)=alpha*G(B)+(1-alpha)*G(A)
    B(C)=alpha*B(B)+(1-alpha)*B(A)

    alpha混合可以实现火光、烟雾、阴影、动态光源等等一切你可以想象的出来的半透明效果，具体思想参考[1]。

    下面给出实现alpha混合的OpenCV代码，取alpha为0.4、0.6时，alpha混合效果如下图。

源代码：

IDE:vc6

unsigned char *data1, *data2, *data3;
int i, j, k;
double alpha;
IplImage *img1, *img2, *img3;
img1 = cvLoadImage("img1.jpg", CV_LOAD_IMAGE_COLOR);
img2 = cvLoadImage("img2.jpg", CV_LOAD_IMAGE_COLOR);
img3 = cvCreateImage(cvGetSize(img1), img1->depth, img1->nChannels);
cvNamedWindow("win1", 1);
cvNamedWindow("win2", 1);
cvNamedWindow("win3", 1);
//  img process
data1 = (unsigned char *)img1->imageData;
data2 = (unsigned char *)img2->imageData;
data3 = (unsigned char *)img3->imageData;
alpha = 0.6;
for(i=0;i<img1->height;i++)
{
    for(j=0;j<img1->width;j++)
    {
        for(k=0;k<img1->nChannels;k++)
        {
            data3[i*img1->widthStep+j*img1->nChannels+k] = (unsigned char)(alpha*data2[i*img1->widthStep+j*img1->nChannels+k] + \
                (1-alpha)*data1[i*img1->widthStep+j*img1->nChannels+k]);
        }
    }
}
//show image and free image
cvShowImage("win1", img1);
cvShowImage("win2", img2);
cvShowImage("win3", img3);
cvWaitKey(0);
cvReleaseImage(&img1);
cvReleaseImage(&img2);
cvReleaseImage(&img3);
cvDestroyWindow("win1");
cvDestroyWindow("win2");
cvDestroyWindow("win3");

unsigned char *data1, *data2, *data3; int i, j, k; double alpha; IplImage *img1, *img2, *img3; img1 = cvLoadImage("img1.jpg", CV_LOAD_IMAGE_COLOR); img2 = cvLoadImage("img2.jpg", CV_LOAD_IMAGE_COLOR); img3 = cvCreateImage(cvGetSize(img1), img1->depth, img1->nChannels); cvNamedWindow("win1", 1); cvNamedWindow("win2", 1); cvNamedWindow("win3", 1); // img process data1 = (unsigned char *)img1->imageData; data2 = (unsigned char *)img2->imageData; data3 = (unsigned char *)img3->imageData; alpha = 0.6; for(i=0;i<img1->height;i++) { for(j=0;j<img1->width;j++) { for(k=0;k<img1->nChannels;k++) { data3[i*img1->widthStep+j*img1->nChannels+k] = (unsigned char)(alpha*data2[i*img1->widthStep+j*img1->nChannels+k] + \ (1-alpha)*data1[i*img1->widthStep+j*img1->nChannels+k]); } } } //show image and free image cvShowImage("win1", img1); cvShowImage("win2", img2); cvShowImage("win3", img3); cvWaitKey(0); cvReleaseImage(&img1); cvReleaseImage(&img2); cvReleaseImage(&img3); cvDestroyWindow("win1"); cvDestroyWindow("win2"); cvDestroyWindow("win3");

参考：

[1] http://dev.gameres.com/Program/Visual/2D/AlphaQiantan.htm