基于SIFT特征配准的多帧迭代盲解卷积算法

Multi-frame iterative blind deconvolution based on the SIFT feature registration

在未配准的情况下,多帧图像解卷积算法不能发挥其利用多帧图像信息互补的优势,针对这种情况,本文将基于SIFT特征配准的算法引入多帧图像解卷积算法中,这一做法克服了各帧影像之间相同位置不同细节像素之间的影响,有效地提高了复原图像的质量。实际观测数据的复原结果证明了本文思想的正确性,配准后图像的复原结果更加清晰,且体现出观测目标更多的细节信息。

Most of the current multi-frame restoration algorithms either utilize the multi-frame data simply by shift-and-add or take no account of the precise registration between frames. With unregistered frames, the complementary information ＂about the identical details can not he effectively used arid the restoration quality is also influenced. Even sometimes with the increase of frames, the result probably turns to be worse. This paper incorporated the SIFT feature registration into the multi-frame deconvolntion, which could overcome the interaction artifacts induced by unregistered images between different detail pixels at the same location and improve the quality of the restored images effeetively. The expriroents proved the superiority of this algorithm. The restored image would be clearer and show more details.