Super-Resolution Technology For Surveillance Video Face Images

The quality of the image affects the accuracy of the obtained information,and it also affects the amount of the obtained information.Therefore,the clarity of image affects the quality of image information directly.High resolution images have an important position in many applications,so the super-resolution algorithm used to improve the image resolution has become a research hot spots in the field of image processing.Compared with the hardware method of improving image resolution,image super-resolution algorithm is low in cost and easy to implement.Because of the advantages of image super resolution,this method has been used in the fields of video,medicine and public security system to improve image resolution and get high resolution image.In this paper,we improve the resolution and the visual effect of surveillance video face image by doing the following work:1.In the light of the problem of monitoring video face image,the technology of surveillance video face image sharpening is proposed.This technology includes improved sparse representation image denoising method and image enhancement method based on the combination of fuzzy theory with guided filter.The experiment shows that our method can not only remove the noise well,but also enhance the color and detail information of the image,and improve the performance of the image super-resolution algorithm.In the technology of surveillance video face image sharpening,we improve the traditional image denoising based on sparse representation,the noise component and the detail component are separated,and the noise component is denoised,so as to retain the details of the image as much as possible.In addition,an image enhancement method based on the combination of fuzzy theory and guided filter is proposed,it can enhance the details of the image,and improve the feature of the surveillance video image:the low contrast,the lack of color information and detail components.2.The image super-resolution algorithm based on convolution neural network is improved.The FERET face database is used as the training set to ensure that the reconstructed face image has a clear outline and the texture is natural.The use of deconvolution avoids pre-processing operations for interpolating to enlarge input images' size.The multi-layer feature mapping layer is used to extract higher level image features and ensure the reconstructed image has better quality.The super-resolution algorithm in this paper can generate images which have richer edge information and better visual effect,and our algorithm is more suitable for monitoring video face images.