总结了图像序列相位关系对超分辨率图像重建效果的影响规律, 对超分辨率图像重建引起的高斯噪声和散粒噪声的放大予以研究.通过叠加不同方差的高斯噪声的图像序列重建实验, 得到结论维纳滤波可以有效地滤除放大后的高斯噪声, 但是超分辨率图像重建后散粒噪声放大成为“波纹”形状的噪声, 传统的中值滤波法不能有效地滤除放大后的“波纹”形状噪声.根据算法处理的是图像序列的特点, 提出了基于图像序列的中值滤波法, 在滤波前通过图像序列的冗余信息判断图像序列上哪些点是受到噪声污染的, 对受到噪声污染的点予以非线性“截断”滤波, 而未受噪声污染的点则不进行滤波, 比传统的中值滤波更有效地滤掉了“波纹”形状的噪声, 而且不会带来图像平滑.

The equation that describes the effect of phase relation of the sequence images on the super resolution image reconstruction was summarized. The amplification of Gaussian noise and shot-noise raised by super resolution image reconstruction was studied. Wiener Filter can restrain the amplified Gaussian noise effectively after super resolution image reconstruction. The shot noise, however, becomes ripple noise after image reconstruction. This brings image confusion because that traditional median filter can not restrain the ripple noise. Considering that the images processed are in sequence, a method of median filter based on sequence images was proposed. The pixels with noise, determined by pleonastic information in sequence images, were eliminated by median filter while the pixels without noise were kept untouched. Median filter based on sequence images restrains noise more powerfully than the one based on single image. It does not bring image confusion and thus improves the quality of the image.