Single image super-resolution(SISR)is a fundamentally challenging problem because a low-resolution(LR)image can correspond to a set of high-resolution(HR)images,while most are not expected.Recently,SISR can be achieve...Single image super-resolution(SISR)is a fundamentally challenging problem because a low-resolution(LR)image can correspond to a set of high-resolution(HR)images,while most are not expected.Recently,SISR can be achieved by a deep learning-based method.By constructing a very deep super-resolution convolutional neural network(VDSRCNN),the LR images can be improved to HR images.This study mainly achieves two objectives:image super-resolution(ISR)and deblurring the image from VDSRCNN.Firstly,by analyzing ISR,we modify different training parameters to test the performance of VDSRCNN.Secondly,we add the motion blurred images to the training set to optimize the performance of VDSRCNN.Finally,we use image quality indexes to evaluate the difference between the images from classical methods and VDSRCNN.The results indicate that the VDSRCNN performs better in generating HR images from LR images using the optimized VDSRCNN in a proper method.展开更多
文摘工业精密制造中,视觉检测设备的成像系统往往景深较小,易产生离焦模糊,严重影响检测效果。针对这一问题,提出了一种针对于均匀离焦图像的盲去模糊网络(Uniform Defocus Blind Deblur Net,UDBD-Net)。首先,提出一种模糊核估计网络,提取离焦模糊的特征,并准确地估计出模糊核;其次,提出一种反卷积网络,通过神经网络学习并估计基于特征维纳反卷积(Feature-based Wiener Deconvolution,FWD)公式中的未知量,更准确地生成去模糊图像的潜在特征;最后,使用一个编解码网络(Encoder-Decoder Net)增强图像的细节,并去除伪影。实验结果表明,该方法在DIV2K和GOPRO图片上的峰值信噪比(Peak Signal to Noise Ratio,PSNR)分别达到31.16 dB和36.16 dB;与目前主流的方法相比,该方法在不显著增加模型推理时间的同时能够复原出更高质量、更自然地去模糊图像。此外,该方法对真实的均匀离焦模糊图像也有较好的去模糊效果,且能够显著提升工业视觉检测算法对于离焦模糊图像的检测效果。
文摘Single image super-resolution(SISR)is a fundamentally challenging problem because a low-resolution(LR)image can correspond to a set of high-resolution(HR)images,while most are not expected.Recently,SISR can be achieved by a deep learning-based method.By constructing a very deep super-resolution convolutional neural network(VDSRCNN),the LR images can be improved to HR images.This study mainly achieves two objectives:image super-resolution(ISR)and deblurring the image from VDSRCNN.Firstly,by analyzing ISR,we modify different training parameters to test the performance of VDSRCNN.Secondly,we add the motion blurred images to the training set to optimize the performance of VDSRCNN.Finally,we use image quality indexes to evaluate the difference between the images from classical methods and VDSRCNN.The results indicate that the VDSRCNN performs better in generating HR images from LR images using the optimized VDSRCNN in a proper method.
