Image Encoding Using Atypical Split Sections for Parallel Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The High Efficiency Video Coding (HEVC) standard's parallel-encoding method, which supports tiles and wavefront parallel processing, suffers from reduced compression efficiency due to interrupted data dependency and limited scalability, especially when splitting images without considering content, leading to deterioration in compression rates.
Innovation Solution
An image encoding method that extracts feature points, performs clustering to determine atypical split sections, and parallel-encodes these sections, allowing for individually determined size and shape, and generates a bitstream with split section information for efficient encoding and decoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If tiles or WPP are used for parallel-encoding, then processing speed and scalability are improved, but compression efficiency deteriorates due to interrupted data dependency
Solution Approach 1:
The picture is divided into multiple split sections based on object boundaries detected through feature point clustering. Each split section corresponds to a distinct object or region, allowing parallel encoding while preserving spatial coherence within each section. This segmentation approach maintains data dependency within objects while enabling parallel processing between different object regions.
Solution Approach 2:
The encoding process applies different processing characteristics to different regions of the picture based on object content. Split sections containing objects with high spatial frequency or complex textures may use different encoding parameters compared to smooth regions, optimizing compression efficiency locally while maintaining overall parallel processing capability.
2Ease of manufacture
If standard rectangular tiling is used, then implementation simplicity is improved, but compression efficiency significantly deteriorates due to ignoring picture content
Solution Approach 1:
Feature point extraction and object boundary detection are performed as preliminary steps before the actual encoding process. This preprocessing identifies optimal split section boundaries based on object locations, enabling content-adaptive splitting without adding complexity to the core encoding algorithm. The split section configuration is determined in advance and passed to the parallel encoding stage.
Solution Approach 2:
The split section configuration is dynamically adapted to the picture content through feature point clustering. Unlike fixed rectangular tiling, the number, size, and shape of split sections vary according to the distribution and characteristics of objects in each picture, optimizing compression efficiency for diverse content while maintaining a unified encoding framework.
3Loss of energy
If WPP encodes first two coding tree units sequentially, then compression efficiency is partially improved, but latency is introduced reducing parallel processing effectiveness
Solution Approach 1:
Each split section is further divided into independent coding tree unit groups that can be processed in parallel. By organizing the encoding workflow around split section boundaries rather than strict sequential coding tree unit processing, the system enables parallel encoding of multiple split sections while minimizing the sequential dependency overhead within each section.
Data Source
AI summary
Provided is an image encoding method including extracting feature points from a picture; generating at least two clusters by performing feature point clustering on the extracted feature points; determining at least two split sections in the picture, the at least two split sections respectively corresponding to the at least two clusters; parallel-encoding the at least two split sections; and generating a bitstream including information about the at least two split sections. A size and a shape of each of the at least two split sections may be individually determined.