Research On Super-resolution Reconstruction Algorithm Of Infrared Cloud Image Based On Learning

Super-resolution is a technique to improve image resolution by signal processing in fixed hardware level.It is an economical and effective means of image resolution enhancement.With this technology,low resolution images can be reconstructed into high-resolution images with richer high-frequency information and clearer texture.In recent years,the super-resolution method based on learning is a hot spot of research.With the vigorous development of meteorological satellite technology,infrared cloud image has played an increasingly important role in weather prediction and monitoring.The resolution of infrared cloud image is lower than that of the visible cloud image.From the perspective of meticulous and comprehensive application,the super-resolution of infrared cloud image is of great value.Therefore,this paper studies super-resolution reconstruction algorithm of infrared cloud image based on this method.The main contents include:First,we introduce the principles,methods and evaluation indicators of superresolution reconstruction algorithm.And then,we introduce the concepts of meteorological satellite and infrared cloud image,and analyze the key characteristics of it.Next,we introduce a super-resolution reconstruction algorithm based on sparse representation for infrared cloud images,and select appropriate parameters suitable for reconstruction of infrared cloud images.Based on the shortcomings of the original method,we propose a hybrid infrared cloud image super-resolution reconstruction method.Experiments show that the improved method combines the advantages of sparse representation and interpolation,and improves the reconstruction speed while improving the reconstruction effect.At last,we introduce the super-resolution reconstruction algorithm of infrared cloud images based on convolutional neural network.Compared with the traditional machine learning method,the deep learning method can optimize the feature extraction operator and the high and low resolution image mapping through the sample training.The feature extraction method has better adaptability and does not have to make the feature extraction methods artificially.The experimental results show that the infrared cloud image super-resolution reconstruction algorithm based on convolution neural network has better reconstruction effect in the edge and flat regions.It shows that the method can effectively represent the rich image features,and is suitable for the reconstruction of large scale infrared cloud images with rich textures and edges.