Perceptual Video Quality Optimization IP

Today's media streaming market demands optimal quality video at the lowest possible data rate, as you need to enhance service users' viewing experience while minimizing platform costs for video transcoding, storage, and delivery. Optimization techniques for video quality play a critical role in streaming applications.

The vendor offers video pre-processing IP solution of AI-powered Perceptual Quality Optimization technology.

The IP solution adaptively filters the input video from a camera or storage before encoding, improving compression efficiency by over 30% while remaining compatible with any codec. Since it preserves perceptual quality in the encoded output, you can reduce service costs without compromising the user’s viewing experience.

Specification

• AXI compliant
  • One AXI4 Lite port with 32-bit address and 32-bit data for the host to communicate with the IP
  • One read and one write channels compliant to AXI4 with 64-bit address and 256-bit data bus
• DNN block
  • Neural Network is specifically designed for in-house PQO(Perceptual Quality Optimization) algorithm
  • Weights of NN are updatable
  • YUV400/420/422 supported
• Performance
  • IP for ASIC achieves 7680x4320@60fps at 600MHz in real-time
• Target devices
  • ASIC
  • FPGA (e.g. Alveo FPGA Cards)

Fully hardwired IP

• Offering high throughput
• No CPU workload required
• AMBA AXI4 compatible

High performance

• Up to 8Kp60 @600MHz in real-time

Codec-agnostic

• IP can be used with any standard video codec

Frame based pre-processor

• Achieving better BD-rate in subjective quality perspective

Robust to coding noise

• Robust to various coding noises in compressed frames