中文摘要：

本文利用阵列相机获取的多幅低分辨率图像合成一幅去噪高分辨率图像,提出一种基于阵列相机的超分辨重建去噪方法.首先,对阵列相机获取图像进行基于卷积神经网络的单幅低分辨率图像的超分辨率重建.其次采用SURF分块匹配方法实现多幅高分辨率图像的配准.最后对配准图像进行基于逐像素多尺度融合的多幅高分辨率图像融合.利用ISO12233分辨率卡进行对比测试,以证明本文方法具有更高图像解析度并减小噪声影响.

英文摘要：

In this paper,a denoising method of super-resolution reconstruction based on array camera is proposed,Which the images captured by the array camera are used to synthesize a denoising high resolution image.First,the super-resolution reconstruction of a single low-resolution image based on convolution neural network is used to process the images captured by array camera.Secondly,the block matching method based on SURF is used to realize the registration of multiple high resolution images.Finally,a multi-scale high-resolution image fusion based on pixel-by-pixel multi-scale fusion is performed on the registration images.Using the ISO12233 resolution card for contrast test,we proved that the method can be used to obtain higher image resolution and reduce the effect of noise.