MSISR : Modified Single Image Super-Resolution Using Relu Based 2D CNN For Satellite Images

In this research, we presented a Modified Single Image Super-Resolution (MSISR) using enhanced very deep super resolution (VDSR). The suggested technique is based on convolutional neural networks and uses up-sampling and residual inputs for training (an essential part of SISR) with a level of 20. The proposed method hybrid fusion of the enhanced bi-cubic method. The proposed MSISR shows better result in terms of peak signal to noise ratio as well as structural similarity index measurement. The proposed method also show good visual outcomes as compare to other previous techniques in terms of butter and resolution. These two factors play a significant role in the outcome analysis of image super resolution (ISR). For the analysis of proposed method use standard data sets like the UC Mecred Field. Data Set are available for the training and testing of the proposed approach. For the simulation of proposed method used well know tool that is matrix laboratory version R2020B.