摘要
旋转图像的质量和性能是需要权衡的一个问题,提出一种细粒度亚像素圆模板算法。通过像素网格状5x5分割选取圆模板的圆心。邻域像素被圆模板覆盖的面积与圆面积的比值Rs,相应像素的颜色值Cs,旋转后像素的颜色值等于Rs和Cs的卷积。实验表明该算法图像旋转质量好于最近邻点插值和高斯模板算法,清晰度和连续性接近双线性插值,计算性能明显优于双线性插值。
The quality and performance of spin image is a problem should to be balanced, a fine granularity sub-pixel circle template algorithm is proposed for this problem. The centre of the circle template is chosen by dividing the centre pixel into 5×5 meshes. Rotated pixel col- our value equals to the convolution of Rs and Cs, here Rs are the ratios of area of circle and areas of neighbourhood pixels covered by circle template, and Cs are the corresponding pixels colours. Experiments show that the quality of image rotation in proposed algorithm is better than that of nearest neighbourhood interpolation and Gaussian template algorithm, its definition and continuity near that of bilinear interpolation,its computation performance is clearly higher than that of bilinear interpolation.
出处
《计算机应用与软件》
CSCD
2009年第5期90-93,共4页
Computer Applications and Software
基金
广东省科技厅科研项目(2003C32401)
关键词
细粒度
圆模板
像素分割
亚像素
高斯模板
图像旋转
Fine granularity Circle template Pixel segmentation Sub-pixel Gaussian template Image rotation