| Summary: | 碩士 === 國立中正大學 === 資訊管理系研究所 === 106 === With advances in technology, images and lifestyle are closely connected and inseparable. A high-resolution image can bring people a great sense of visual perception. Moreover, high quality and clear informative details can not only reduce the demands of image preprocessing but achieve the better effects.
In this study, we apply two kinds of convolutional neural networks which are SRCNN and DCSRN to super-resolution image reconstruction. Furthermore, reconstructed images in this study are specific to remote sensing images. We focus on enhance the quality of reconstructed remote sensing images and reduce the time of training and executing model. First, we adjuste the parameters of convolutional neural networks and reduce the size of the dataset to train the model. Second, we evaluate the reconstruction results of different models with the human eyes and evaluate indicators. Last but not least, we discuss the quality of the reconstructed remote sensing images and the time spent on it in order to provide an objective reference for researchers who need to research high-resolution remote sensing images.
|
|---|
