How HEVC H.265 Improves Upon MPEG-4 AVC H.264

This article explores how High Efficiency Video Coding (HEVC/H.265) advances beyond its predecessor, MPEG-4 AVC (H.264). We examine the key technological enhancements—including larger coding tree blocks, expanded intra-prediction modes, and improved parallel processing—that collectively allow HEVC to deliver up to 50% better data compression while maintaining the same visual quality.

Larger and More Flexible Coding Blocks

MPEG-4 AVC relies on “macroblocks” with a fixed maximum size of 16x16 pixels to analyze and compress video frames. HEVC replaces these with Coding Tree Units (CTUs), which can scale up to 64x64 pixels. These larger blocks are highly flexible and can be dynamically subdivided into smaller coding units based on the detail in the scene. For high-resolution video like 4K and 8K, where large portions of the screen often contain similar colors or textures (such as a blue sky), these larger blocks allow HEVC to compress data far more efficiently than AVC.

Enhanced Intra-Frame Prediction

Intra-frame prediction reduces spatial redundancy within a single frame. While MPEG-4 AVC supports only 9 directional prediction modes, HEVC expands this to 33 directional modes, alongside planar and DC modes, totaling 35. This dramatic increase in prediction directions allows HEVC to more accurately predict pixel trends and edges within a frame, resulting in sharper image details and a significant reduction in the amount of data needed to represent complex textures.

Advanced Motion Vector Prediction (Inter-Frame)

To reduce temporal redundancy between consecutive frames, HEVC introduces superior motion estimation and compensation techniques. HEVC utilizes Advanced Motion Vector Prediction (AMVP) and Merge Mode. These tools allow the encoder to reuse motion vectors from neighboring spatial blocks or preceding frames, minimizing the metadata required to describe movement. This results in smoother motion tracking with far fewer bits.

Sample Adaptive Offset (SAO) Filter

While both standards use a deblocking filter to smooth out pixelated edges, HEVC introduces an additional in-loop filter called Sample Adaptive Offset (SAO). SAO analyzes the reconstructed video signal and applies offset values to better match the original source picture. This step reduces ringing artifacts and banding, particularly in areas with subtle color gradients, leading to a much cleaner final image.

Built-in Parallel Processing Support

MPEG-4 AVC was not designed with multi-core processors in mind, making parallel processing difficult. HEVC addresses this by introducing Tiles and Wavefront Parallel Processing (WPP). These features allow a single video frame to be divided into independent grid-like regions or sequential rows that can be encoded and decoded simultaneously across multiple CPU cores. This architecture makes HEVC highly optimized for modern hardware, enabling the real-time processing of high-bitrate video streams.