Images constitute a significant portion of data transfer between end-user devices and servers on many websites, while simultaneously driving the energy consumption of display screens - often the most power-intensive component in end-user devices.
This work employs Super Resolution (also known as Super Scaling), a technique commonly used in computer gaming that enables low-resolution images to be upscaled to high-resolution equivalents, thereby requiring only the transmission of smaller, low-resolution files.
A Deep Neural Network was trained to perform this image transformation and integrated into an experimental browser based on Servo, demonstrating how neural networks can effectively reduce energy consumption associated with image transfer.
Early analysis into the energy consumption of the network compared to the high-resolution data transfer shows promising results.