一种改进的字典学习的教室图像超分辨率重建方法

An Improved Dictionary Learning Super-resolution Reconstruction Method for Classroom Images

目前,教室的成像因受设备性能低和环境复杂的影响,会出现教学环境下对师生认识不全的情况。为了充分利用图像信息,全面细致地了解教学情况,文章提出一种改进的字典学习的教室图像超分辨率重建方法。通过采用字典学习算法训练自构的教室图像数据集得到对应的低秩字典和稀疏字典,使用训练的两个字典重建训练集图像,再参与训练,得到残差字典,然后运用训练得到的三个字典重建低分辨率图像,最终得到高分辨率图像。将提出的算法与几种经典算法进行对比实验,可视化和量化结果均表明,提出的算法在PSNR和SSIM上都获得了显著的提升。

At present,the imaging of classrooms is affected by low equipment performance and complex environments,resulting in incomplete understanding of teachers and students in the teaching environment.In order to fully utilize image information and comprehensively and meticulously understand the teaching situation,this paper proposes an improved dictionary learning super-resolution reconstruction method for classroom image.By using dictionary learning algorithms to train a self constructed classroom image dataset,corresponding low rank and sparse dictionaries are obtained.The two trained dictionaries are used to reconstruct the training set images,and then participate in training to obtain residual dictionaries.Then,the three trained dictionaries are used to reconstruct low resolution images,ultimately high-resolution images are obtained.Comparative experiments are conducted between the proposed algorithm and several classic algorithms,and both visual and quantitative results show that the proposed algorithm achieved significant improvements in both PSNR and SSIM.