基于差分进化的插值置换混叠图像盲分离

Blind Separation Method for Interpolation Permuted Alias Image Based on Differential Evolution

针对插值置换混叠图像提出了一套完整的盲分离方案。根据插值图像像素点之间存在的相关性,通过对置换混叠图像分块进行有限差分,检测插值图像与原图像经过有限差分后的值之间的差距,设定适当的阈值,将其分开。为获得最优阈值,用差分进化算法进行优化,获得最优阈值。根据阈值将图像二值化,进而将置换图像分离出来。实验结果表明,该方法比阈值法的鲁棒性强,对经过不同插值方式的置换混叠图像进行盲分离都有较好的效果。

To deal with interpolation permuted alias image, a complete blind separation method is proposed in this paper. According to the correlation between the pixels of the interpolated image, the permuted alias image is divided into sub-blocks which will be processed by finite-difference. Then, each of the processed sub-blocks is assigned a threshold by measuring the difference between the values of the interpolated image after finite-difference and the original image. Next, the differential evolution is performed to obtain the optimal threshold vector for thresholding the image into binary one and finally the permuted image can be separated. The experimental results demonstrate that, compared with the thresholding approach, this algorithm indicates a stronger robustness and can obtain better performance of btind separation for interpolation permuted alias images processed by various interpolation methods.