摘要
基于拉普拉斯金字塔理论,通过外推的方式预测高频分量,提出多尺度的图像增强算法。对图像拉普拉斯残差的过零点提取图像边缘,并分别对水平分量和垂直分量的边缘进行增强。通过灰度值门限划分高对比度区和低对比度区,分别进行不同增幅倍数的增强,实现不同对比度增幅的多尺度增强;通过边缘半径指定多尺度增强范围,解决现有算法中只增强过零点邻居像素灰度值的缺陷,实现边缘对比度的自然过渡。仿真表明,所提算法较现有算法具有更好的PSNR性能,并具有更好的可调性。
A muhi-scale image enhancement algorithm is proposed, based on Laplacian pyramid theory and by predicting the high-frequency components via extraplating mode. The edges of image are abstracted by using the zero-crossing points of Laplacian residual. The horizontal and vertical components of edges are enhanced respectively. A gray-level threshold is set to distinguish the high and low contrast regions which are enhanced by different amplifications. So multi-scale amplification enhancement is implemented. The enhanced range around edges is set by edge radius which implements the natural transition of contrast. It solves the enhancement limitation of recent method which only enhances the neighbor of zero-crossing points. The simulation results indicate that the proposed algorithm is superior to recent method in terms of PSNR and fairly adjustable.
出处
《通信技术》
2017年第3期455-458,共4页
Communications Technology
关键词
多尺度图像增强
拉普拉斯金字塔
高频分量外推
边缘半径
multi-scale image enhancement
Laplacian pyramid
high-frequency component extrapolation
edge radius
