一种自适应协方差的低复杂度图像放大方法

发布时间：2018-07-03 06:37

本文选题：图像放大 + 自适应协方差　；参考：《电视技术》2017年Z2期

【摘要】：对于图像放大技术而言,重要的就是要权衡到图像质量以及计算复杂度。传统的基于线性或三次样条插值的方法会带来图像模糊和锯齿边缘等失真,为了解决这一问题,人们提出一种基于迭代和学习的算法,但是这种方法带来了很高的计算复杂度。综合以上几点本文提出了一种基于自适应协方差的图像放大方法(adaptive covariance-based edge diffusion,ACED)。该方法能很好地权衡图像放大性能和复杂度之间的关系。在这种方法中,提出了一种联合边缘判别准则,并自适应选择扩散模板来估计局部协方差系数,以高效的减少图像放大带来的失真。实验结果表明,所提出的方法在主观质量和客观质量上都有很大的提升,同时也具有较低的计算复杂度。
[Abstract]:For image amplification, it is important to balance image quality and computational complexity. Traditional methods based on linear or cubic spline interpolation can bring distortion such as image blur and sawtooth edge. In order to solve this problem, an iterative and learning-based algorithm is proposed, but this method brings high computational complexity. In this paper, we propose an adaptive covariance-based image amplification method (adaptive covariance-based edge diffusion-ACED). This method can balance the relationship between image magnification performance and complexity. In this method, a joint edge discriminant criterion is proposed, and a diffusion template is chosen to estimate the local covariance coefficients to reduce the distortion caused by image amplification. The experimental results show that both subjective and objective quality are greatly improved and the computational complexity is low.
【作者单位】：上海大学通信与信息工程学院;
【基金】：国家自然科学基金项目(61271212) 国家863计划项目(2015AA015903)
【分类号】：TP391.41

【相似文献】