Imagination Technologies and Visidon Oy join forces to drive the transition to deep learning-based super-resolution for embedded applications in the mobile, digital TV and automotive markets. With this artificial intelligence (AI) technology, users can increase the resolution of low-resolution images and videos to 4K and 8K through advanced algorithms. The IMG Series4 NNA is able to complete such demanding processes in real-time with higher efficiency under the stringent power and thermal requirements of the SoC.
Imagination's IMG Series4 Neural Network Accelerator (NNA) brings powerful computing performance and leading energy efficiency to advanced AI image processing software through its Tensor Tiling technology. Combined with Visidon's expertise in deep learning video enhancement networks, the solution enables OEMs to apply cutting-edge AI enhancement techniques to image inpainting and colorization, restoring low-resolution source data through up-sampling For high-resolution output, the video transmission bandwidth is greatly reduced. .
Matti Niskanen, CTO of Visidon, said: “The two pillars of the Visidon solution are energy efficiency and high performance. We were very surprised to see how easily our deep learning network could run on Imagination’s IMG Series4 NNA, while also being The overall energy efficiency maintained when processing high computing workloads is impressive. We hope that the two parties can further deepen their cooperation and jointly open up the future of AI super-resolution technology."
Gilberto Rodriguez, Director of Product Management at Imagination, said: “Both Imagination and Visidon build technology with guaranteed performance and power consumption. We are very excited to be working with Visidon to advance the development of advanced AI image processing technology, enabling OEMs to use leading IP and software provide super-resolution applications. It is hoped that the collaboration between our two companies will bring new opportunities for innovation in mobile, digital TV and automotive SoCs.”
The IMG Series4 AI computing engine uses Tensor Tiling technology to run the convolutional neural network, the main algorithm in the Visidon super-resolution solution portfolio, when the system bandwidth is low and high inference speed is required.
Visidon has developed a deep learning-based super-resolution network to upscale and fix the resolution of video transmissions to the 4K (2160p) and 8K (4320p) resolutions required by modern monitors. Visidon super-resolution technology increases video frame size while preserving image detail, achieving ultra-high output quality superior to industry standard upscaling algorithms. Visidon generally avoids the increase in noise by combining super-resolution with noise reduction techniques, while preserving realistic detail and clarity.
