Research On The Application Of Multifunctional Sub-network Model In Image Super-resolution Reconstruction

The resolution of an image is an important indicator to measure the quality of an image.Compared with low-resolution images,high-resolution images have more extensive application requirements in many fields such as medicine,remote sensing,and security.However,limited by hardware conditions,many imaging devices can only output low-resolution images.How to break the imaging limit of hardware facilities at the software level to enhance the resolution of images is of great research significance,and image super-resolution reconstruction is such a technical method that can obtain high-resolution images from low-resolution images.Compared with the traditional algorithm,the existing image super-resolution reconstruction method based on deep learning has greatly improved in terms of algorithm performance and reconstruction result.Based on deep learning,this paper analyzes the image degradation model from the high-resolution original image to the low-resolution observation image,and uses the deep learning idea to build a multi-function sub-network model.The low-resolution image contains noise and high resolution In-depth study on the issue of rate reconstruction.The main research results of this article are as follows:1.In view of the noise contained in the low-resolution image or the noise caused by the image upsampling process from low resolution to high resolution,this paper proposes an image denoising restoration algorithm based on the enhanced codec network.In this algorithm,multiple enhanced codec units are connected in a dense connection to form an enhanced codec denoising neural network.The multiple enhanced codec units are used to cyclically remove noise from the noisy image input,and finally obtain a satisfactory clean image.The test comparison experiment is carried out on the general noise data set,and the results prove that the image denoising effect of the algorithm proposed in this paper is significantly better than other denoising restoration algorithms.2.The existing image super-resolution reconstruction method based on convolutional neural network has a single focus in the reconstruction process,and it is easy to ignore the impact of complex environments on image imaging.This paper proposes an image super-resolution reconstruction algorithm based on multi-function sub-network.The algorithm first constructs a functional sub-network,one sub-network is a deep convolutional network structure,and the other is a simplified structure of the enhanced codec network proposed above.The two sub-networks are given different functions: the characterization of image texture details and structure information and the degradation of image noise,both of which operate independently and synchronously.Then,the sub-networks of different functions are connected and fused to obtain high-level abstract image features.Finally,a high-resolution image is reconstructed by combining the high-level abstract features of the image and the initial features of the image.The high-resolution images reconstructed by the functional sub-network not only have clearer texture details but also have higher objective evaluation index values.Experimental results show that the proposed algorithm is superior to the current best image super-resolution reconstruction algorithm in both objective evaluation indicators and subjective visual effects.