Research And Design Of Adaptive Image Denoising System Guided By Image Quality Assessment

Noise are easily introduced in the process of image acquisition,transmission and storage,which leads to the degradation of image quality.Thus,image denoising is necessary because high-quality images are the basis of advanced vision tasks.The sensitive edge area of the human eye is prone to over-smoothing during denoising.How to obtain images that consistent with human subjective preference by simulating the human visual system(HVS)is one of the research hotspots in the field of image denoising.This paper proposes to adaptively remove image noise guided by image quality assessment to obtain images with higher visual quality.By studying the characteristics of HVS,a no-reference image quality assessment method based on dual-quality map is researched and designed.Aiming at the problem of distortion types and content diversity in multiple scenes,this method uses a multiscale generative network to predict dual-quality map and use it as intermediate target,which removing redundant information that is irrelevant to image quality.Meanwhile,spatial attention and channel attention mechanisms are introduced into the generative network to capture the distortion information that is of interest to the human eye.The deformable convolution is used in the quality prediction network to replace the deep conventional convolution,and the adaptive adjustment of the receptive field is realized by predicting the offset of the sampling point.The experimental results show that the method proposed in this paper has high evaluation accuracy on both synthetic and authentic distorted image datasets.The SROCC and PLCC on the MDID2017 dataset reach 0.968 and 0.966,respectively,which are highly consistent with the subjective evaluation results.Guided by the proposed NR-IQA method based on dual-quality map,an image denoising technique with adaptive parameter adjustment is proposed.The standard deviation of the Gaussian function is dynamically adjusted according to the image quality result and the distortion type detection probability,then calculates the filter coefficient in combination with the edge direction to prevent the over-smoothing phenomenon of the denoised image.Finally,an image denoising prototype system is designed and realized,which includes filter coefficient calculation,adaptive Gaussian filtering,data buffering and output display units.The image denoising system has been experimentally tested on the CBSD68 public noise dataset.The experimental show that after adding the image quality assessmentguided standard deviation adaptive adjustment algorithm,compared with the conventional Gaussian filter denoising method,the PSNR and SSIM values are improved by 9.51% and 6.61% respectively,the MSE value is reduced by 43.95%.The results of the system have good performance in both subjective visual perception and objective index,indicating that the denoising system can effectively remove the noise in images.