For the past ten years, MPEG-2 encoders based on ASIC chips have assisted in the growth of digital video services. These chips are reasonably fast, well understood and mature in their application. However, ASIC based encoding products require long development cycles for new generations of chips, and are limited in their ability to deliver optimal combinations of image quality and low bit rates.

Unlike ASIC encoders, EGT utilizes high-speed Digital Signal Processors. This fresh approach delivers several critical advantages for operators:

• Rapid encoder improvements—development cycles are measured in weeks and months, not years
• Unmatched video quality—EGT applies HVS (human visual system) models for bit allocation, and redefines the distortion-masking capability of HVS within spatial and temporal domains—simply put, EGT’s products compress video so it looks normal to end users, at bit rates lower than any ASIC-based competitor
• Absolute flexibility—Because the algorithms are implemented on programmable processors, all aspects of the encoder can be optimized and nothing is locked behind ASIC hardware

  How it Works
  Video compression systems execute a series of bit allocation decisions. Encoders typically combine predefined thresholds, parameters and models for these decisions.

  Unlike typical encoders, EGT’s Encoders utilize an innovative adaptive video coding model called PicturePerfect. PicturePerfect continuously adapts the operation of the encoder to the content being coded by optimizing components such as the scene-change detector, frame bit-allocator, motion estimator and quantizer selector.

  Components of the PicturePerfect algorithm include:

  • Adaptive Hierarchical Motion Estimator – Adaptively selects the number of hierarchy levels to trade-off the bits used to code motion vectors and residuals
  • Look-Ahead Complexity Estimator – Estimates frame complexity and appropriately allocates bits across frames and GOPs (Group of Pictures) for uniform temporal quality before actual coding
  • Quality Optimized Macroblock Coding Type Estimator – Estimates the optimal coding type for a macroblock (intra, non-intra, forward, backward, bi-directional)
  • Parametric Frame Coder – Collects statistics from all macroblocks in a frame and uses a dynamic parametric model for appropriate bit allocation per macroblock