快速离散化双线性插值算法

Fast discrete bilinear interpolation algorithm

双线性插值算法在数字图像处理中有广泛的应用。它具有比最近邻点法更好的连续性,但计算速度慢。为提高其计算速度,提出了离散化双线性插值算法。把像素分割成子像素,用双线性插值函数计算子像素中心坐标处的像素值,可以得到每一个子像素的卷积函数C,用C来代替双线性插值函数。理论分析表明离散化算法减少了加法和乘法的浮点运算次数。通过图像旋转实验表明该算法与双线性插值算法相比计算速度有较大的提高,连续性略有损失,图像质量基本相同。

Bilinear interpolation is widely used to process digital image. Its continuity is better than Nearest neighbor interpolation, but is time-consuming. In order to speed up, a discrete bilinear interpolation algorithm is proposed. First, pixel is cut into 3x3 sub-pixel. Then the center value of each sub-pixel with bilinear interpolation function is calculated, we can get a convolution function C related with eachsub-pixel, the function C can be a substitute ofbilinear interpolation. Theoretical analyses show that the discrete algorithm decreases the times of floating addition and floating multiplication. Experiments of image rotation show that the discrete algorithm improve the performance more greatly than that of bilinear interpolation. The quality of image rotated by the discrete algorithm is almost equal to that by the bilinear interpolation except for a little loss of continuity.