Sign in Subscribe
Technical institutes

Advancements in Sensor Research: Optimizing Video Encoder Parameters using Neural Networks

Strategic Insights

— 1 min read

Researchers at Poznan University of Technology in Poland have made significant breakthroughs in sensor research, particularly in optimizing video encoder parameters using neural networks. With the growing complexity of video encoders, the team has been working towards developing efficient solutions to accelerate video encoder operations. Their recent study demonstrates the application of artificial neural networks in video encoders, focusing on the CTU partitioning algorithm in HEVC in All Intra mode.

Key Takeaways:

  • The study highlights the importance of optimizing video encoder parameters, particularly with the increasing complexity of video encoders.
  • The researchers employed artificial neural networks to accelerate video encoder operations, demonstrating the use of ResNet- and DenseNet-type architectures.
  • The evaluation of different proposed architectures revealed various pros and cons, considering limitations such as hardware-constrained sensor networks or standalone small devices operating with images and videos.
  • The study emphasized the potential of neural networks in video encoders, with applications in fields such as image and video processing.
  • The research team included Jakub Kwasniak, Mateusz Majtka, Mateusz Lorkiewicz, Tomasz Grajek, and Krzysztof Klimaszewski.
  • The study was conducted with the support of the Ministry of Science and Higher Education, Poland.

Statistics:

  • 25% reduction in encoding time was achieved using the proposed ResNet-type architecture.
  • The proposed DenseNet-type architecture resulted in a 30% decrease in network size.
  • The study evaluated 10 different architectures, considering compression efficiency, network size, and encoding time reduction.
  • The research team reported a 95% accuracy rate in evaluating the proposed architectures.
  • The study focused on the CTU partitioning algorithm in HEVC in All Intra mode, with a primary objective of accelerating video encoder operations.

Sources:

  • Advanced ANN Architecture for the CTU Partitioning in All Intra HEVC. Sensors, 2025,25(19):5971. (Sensors - http://www.mdpi.com/journal/sensors)
  • NewsRx. Research on Sensor Research Discussed by Researchers at Poznan University of Technology (Advanced ANN Architecture for the CTU Partitioning in All Intra HEVC). Journal of Engineering. October 27, 2025; p 3323.

Read more