Image Denoising And Classification Method Based On Deconvolutional Network

Image representation is an important core issue in digital image processing and analysis,and it is also a hot topic in the field of computer vision today.Common image representation methods generally cannot fully consider the content contained in the image,and it is difficult to meet the requirements of high-level and low-level computer vision tasks,such as image denoising and classification.Image representation based on image features can solve this problem well,but such methods are usually not used for image reconstruction tasks.Therefore,high-level and low-level computer vision tasks based on image feature representation are a subject worthy of study and challenge.This paper designs and implements an image denoising and classification model based on deconvolutional network.The experimental results show that this method has certain advancement and potential application value.Firstly,the theory and implementation of deconvolutional network are introduced.Then,the denoising network filter and feature map are used to reconstruct the noise image and an image denoising method is implemented.Finally,The feature maps derived from the product network are encoded into feature vectors and the SVM classifier is used to implement the image classification method.The experimental results show that the deconvolutional network has good feature learning ability and good modeling ability for low-level image features in the image.The main work of this article has the following aspects:1.This paper first systematically reviews the deconvolutional network and its learning process.Since image representation based on image features can meet the requirements of highlevel and low-level computer vision tasks,it is usually not used to implement image reconstruction and denoising.This paper designs and implements an image denoising and classification model based on deconvolutional network.2.The image denoising method is designed based on the deconvolutional network derived network filter and feature map to achieve image denoising.Experimental results show that the denoising method in this paper has certain advantages over other commonly used denoising methods.3.Difficulties that image features represent in image classification are optimized.The deconvolutional network is used to optimize image feature representation and an efficient image classification model is constructed.Using the method designed in this paper,Caltech-101 image data is classified,and 66.8% classification accuracy rate is achieved.