Video Encoder Frame Buffer Memory Management

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video encoding and decoding systems require substantial memory to store frame buffers for motion compensation, leading to increased hardware costs and power consumption, while also complicating random access to prediction blocks.

Innovation Solution

A method for motion estimation or compensation that involves storing a reference picture in a frame buffer, overwriting a subset of it with a current picture, and protecting specific blocks of the reference picture to prevent data loss, allowing for efficient reuse and reduced memory requirements.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If frame buffers are allocated to store reference pictures for motion compensation, then motion estimation accuracy is improved, but memory storage requirements increase

Engineering Contradiction:
Improvemotion estimation accuracyVSAvoidmemory storage requirements
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The frame buffer is segmented into multiple banks (first bank and second bank) that can be independently managed. Reference pictures are stored in one bank while current pictures are stored in another, allowing parallel access and reducing the need for large contiguous memory blocks. This segmentation enables the system to maintain adequate reference picture storage without requiring excessive total memory capacity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a temporal dimension to memory management by implementing a wrapping mechanism where frame buffers循环利用 (reuse) after their reference period expires. Instead of allocating permanent storage for all reference pictures, the system reuses buffer space over time, transforming the memory management problem from a static spatial allocation issue to a dynamic temporal reuse strategy, thereby reducing peak memory requirements.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Ease of operation

If large frame buffers are used to store multiple reference pictures, then random access to prediction blocks is improved, but hardware complexity increases

Engineering Contradiction:
Improverandom access to prediction blocksVSAvoidhardware complexity
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The system implements dynamic buffer management where the frame buffers can be switched between different operational states (reference picture storage, current picture storage, and reuse). The wrapping mechanism dynamically allocates and deallocates buffer space based on the current processing stage and reference picture requirements, allowing the hardware to adapt to varying access patterns without requiring permanently allocated large memory structures for all possible scenarios.

Inventive Principle:
Principle #15Dynamics

3Measurement precision

If reference pictures are stored in frame buffer, then motion compensation accuracy is improved, but power consumption increases

Engineering Contradiction:
Improvemotion compensation accuracyVSAvoidpower consumption
Core Design Contradiction:
Measurement precisionVSUse of energy by stationary object

Solution Approach 1:

The patent implements a reference picture buffer management strategy where reference pictures are discarded from the frame buffer after their reference period expires and the buffer space is recovered for reuse. This selective discarding and recovery mechanism ensures that only necessary reference pictures are maintained in memory at any given time, reducing the overall memory occupancy and associated power consumption while maintaining motion compensation accuracy for the required reference period.

Inventive Principle:
Principle #34Discarding and recovering

Data Source

PatentUS7925120B2Methods of image processing with reduced memory requirements for video encoder and decoder
Publication Date: 2011.04.12 XUESHAN TECH INC
  • US7925120B2 patent drawing
  • US7925120B2 patent drawing
  • US7925120B2 patent drawing

AI summary

A method of performing image processing includes storing a reference picture into a frame buffer starting at a first starting address; overwriting at least a subset of the reference picture by storing at least a portion of a current picture into the frame buffer starting at a second starting address; protecting a plurality of blocks of the reference picture from being lost while storing each block of the current picture in the frame buffer; and performing a motion compensation procedure for a particular block of the current picture in the plurality of blocks of the reference picture that are being protected.