Parallel Image Decoding with Synchronization State Selection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Decoding large compressed image files is often a time-consuming process due to the serial nature of bit-sequences, making it difficult to decode different parts of the sequence in parallel without additional information, and existing methods rely on unpredictable natural resynchronization, which is not reliable for error-free decoding.

Innovation Solution

A method to determine a synchronization state in a sub-sequence that allows for parallel decoding of compressed image files by selecting a set of states that ensures correct decoding of the remainder of the sub-sequence, reducing storage needs and enabling efficient and predictable synchronization, thereby allowing multiple sub-sequences to be decoded simultaneously.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If parallel decoding of sub-sequences is implemented, then decoding speed is improved, but synchronization reliability deteriorates due to unpredictable natural resynchronization

Engineering Contradiction:
Improvedecoding speedVSAvoidsynchronization reliability
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent applies preliminary action by pre-calculating and storing synchronization states in a lookup table during the encoding phase. These synchronization states are determined in advance and stored with their corresponding positions in the compressed data stream. During parallel decoding, the decoder simply queries this pre-computed table to find valid starting positions, eliminating the need for unpredictable natural resynchronization and ensuring reliable synchronization across multiple parallel decoding threads.

Inventive Principle:
Principle #10Preliminary action

2Productivity

If block boundaries are used for parallel decoding, then decoding can be parallelized, but erroneous decoding occurs until synchronization is achieved

Engineering Contradiction:
Improveparallel decoding capabilityVSAvoiddecoding accuracy
Core Design Contradiction:
ProductivityVSManufacturing precision

Solution Approach 1:

The patent introduces an intermediary mechanism in the form of a synchronization state lookup table that acts as a mediator between the compressed data blocks and the decoding process. Instead of directly decoding from arbitrary block boundaries (which causes errors), the decoder uses this intermediary table to identify and select only those block boundaries that correspond to valid synchronization states. This intermediary layer ensures that parallel decoding blocks start at correct positions, eliminating erroneous decoding while maintaining parallelization benefits.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Device complexity

If natural resynchronization is relied upon, then parallel decoding is simpler to implement, but the decoding point is unpredictable and error-prone

Engineering Contradiction:
Improveimplementation simplicityVSAvoidsynchronization point accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent applies preliminary action by pre-calculating and storing synchronization states in a lookup table during the encoding phase. These synchronization states are determined in advance and stored with their corresponding positions in the compressed data stream. During parallel decoding, the decoder simply queries this pre-computed table to find valid starting positions, eliminating the need for unpredictable natural resynchronization and ensuring reliable synchronization across multiple parallel decoding threads.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP2975771B1A method for selecting starting positions in parallel decoding of a compressed image
Publication Date: 2020.12.30 PHASE ONE
  • EP2975771B1 patent drawingFigure 1~2
  • EP2975771B1 patent drawingFigure 3~4H
  • EP2975771B1 patent drawingFigure 5~6

AI summary

The invention provides a method to divide a sequence of symbols being encoded as variable lengths prefix-free code-words in a compressed image file into sub-sequences which can be decoded without referring to any other sub-sequence. Hence, it is possible to divide the original data into several sub-sequences which can be decoded in parallel and assembled to form the fully decoded sequence at the end. In particular, the invention relates to a way to determine a synchronization state in a sub-sequence allowing correct decoding of the remainder or the sub-sequence.