分析了复眼图像超分辨率重构结果图像中噪声来源及主要表现形式，为优化重构结果，提出一种基于Wiener-Butterworth（W-B）滤波器的重构噪声消除方法。通过叠加不同数量高斯噪声的低分辨率图像重构实验，研究了该滤波器对算法引入噪声和原始图像噪声的滤除效果，并在频域对实验结果作了相关分析；验证实验验证了该滤波器在噪声类型和大小均未确定时改善图像质量的有效性。实验结果表明：针对算法引入噪声，该滤波器使重构结果的PSNR提高4 dB；当施加均值为0、方差为0.01的高斯噪声时，滤波后重构结果PSNR提高6.9 dB。W-B滤波器能有效消除复眼超分辨率重构图像中的噪声和方块效应、振铃效应和梳边效应，改善重构图像质量。%The sources and manifestation of noise in result of the compound eye super resolution reconstruction are analyzed. In order to optimize the results of the reconstruction, a filtering method based on the wiener-Butterworth filter is proposed to eliminate the impact of noise in the processing of reconstruction. The performance of the wiener-Butterworth filter for eliminating the introduction noise of algorithm and the original noise is analyzed through the simulation, and the result of reconstruction is analyzed in frequency domain. The wiener-Butterworth filter can improve the quality of the reconstruction distinctly when the type and number of noise can not be determined. The result shows that the wiener-Butterworth filter increases the PSNR of the reconstruction 4 dB when the low resolution images with no noise and increases the PSNR of the reconstruction 6.9 dB when the low resolution images with Gaussian noise with the mean is 0 and the variance is 0.01. The Wiener-Butterworth filter can eliminate the blocks effect, comb effect and ringing effect of the reconstruction effectively and improve the quality of the reconstruction.
展开▼
