Using Image Upscaling Methods in Digital Platforms to Reduce Internet Usage
Willy Lau, Josia Sean Audric Santoso, Ignatius Ronald, Eko Setyo Purwanto, Azani Cempaka Sari · 2022 4th International Conference on Cybernetics and Intelligent System (ICORIS) · 2022
Image scaling has played an essential role today, especially in file sharing, such as on social media. Image scaling is a process for changing the size of a digital image. There are a lot of methods in image scaling that are being used in digital platforms. The standard image scaling being used is the upscaling method. Image upscaling also requires a lot of time to process, making it slower to finish upscaling the photos. This paper has proposed an optimized upscaling method where the upscaling process only focuses on the object in the images. By using the Object Detection AI, the Image Upscaling Method will only focus on the object inside the picture. Theoretically using image upscaling can reduce the time needed to finish the upscaling process. Before implementing this object-focused AI in the image upscaling method, there will be done a comparison of several upscaling methods to determine which upscaling method has the best quality. The results of comparison between some image upscaling method shows that the waifu2x has the best image quality. Not only the image quality, this waifu2x also has one of the smallest image size produced by the algorithm compared to the others. In this experiment, there is a limitation in experimenting the algorithm because using waifu2x requires CUDA Hardware Acceleration by Nvidia. Without using CUDA, the time needed to finish the upscaling process requires longer time.