Combining super-resolution algorithm (gaussian denoising and kernel blurring) and comparing with camera super-resolution

A good Super Resolution (SR) algorithm is one of the key successes to filter frequency that creates noise to a picture. Previous research that has published was concluded the Camera SR is the best algorithm to filter this frequency based on their Peak Signal to Noise Ratio (PSNR) and Mean Square Error (MSE) results. However, the current approach to achieving high resolution have not yielded enough signal to filter unwanted pixel. Hence, there is a need to find a better approach to those leads to higher resolution through lower noise reduction. To fulfill this need, this thesis proposed to utilize two proven SR algorithms; Gaussian Denoising and Kernel Blurring. This thesis will not only be obtaining these two existing algorithms in a stand-alone form buthence the combination of them (two combinations) will also be obtained as the new possible algorithms that can be utilized to filter frequency that create noise to a picture. To reach the research
objective, the method that will be used is by training a total of four algorithms one by one to a public data set that contains 200 pictures and gets the PSNR and MSE results of each algorithm. Comprehensive experimental results show that all those four SR algorithms outperform previous SR algorithms in commonly used data set with variously higher PSNR by 21% and lower MSE by 5%.