基于奇异值分解与重采样的图像缩放

SVD and resampling based image resizing

针对传统方法的缺点,本文提出了一种新的图像缩放算法.经典插值缩放方法忽略了图像纹理之间的突变特性,因而导致高频信息的退化.同时在进行高倍数图像放大的时候,使用该方法容易造成马赛克现象的出现.基于小波变换的图像缩放方法只能够进行原始图像偶数倍的放大,且放大效果并不一定理想.本文提出的图像放大算法,联合使用奇异值分解和重采样操作,对图像进行放大处理,不但可以克服边缘模糊化以及马赛克现象的产生,同时可以进行任意倍数的图像放大.大量实验结果表明,较传统方法而言,本文方法不仅具有良好的视觉效果,同时峰值信噪比以及灰度绝对偏差等客观评价标准也达到一定的性能指标.

A novel algorithm based on singular-value decomposition and resampling is proposed in the paper. Most resizing methods use interpolation in a way or another. Interpolation methods are conventional in that an interpolation function is applied indiscriminately on the whole image. It＇ s fast but since frequencies near cutoff frequency is attenuated, image blurring and smoothing may be brought. Meanwhile, because of the intrinsic character, wavelet based interpolation can not resize image arbitrarily. An algorithm based on singular-value decomposition and resampling is used in the paper to resize image arbitrarily. Experiments show that the result is satisfying.