Motion Analysis via Video Compression Code Stream Macro Blocks

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video motion analysis methods in security monitoring systems are limited in detecting motion in various coded streams and scenarios, particularly failing to accurately detect slowly moving objects and adapt to different coding conditions, leading to inaccurate results and high computational requirements.

Innovation Solution

A motion analysis method based on video compression code streams that extracts and weights the bit number of macro blocks, calculates differences, and uses threshold determination to detect moving objects without needing to calculate inter macro block pictures, reducing computational load and improving adaptability.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If video motion analysis is performed by decoding compressed code streams and counting pixel features of decoding pictures, then motion analysis can be performed on decoded video data, but a great deal of computation and larger storage space are needed, making real-time processing difficult to achieve

Engineering Contradiction:
Improvemotion analysis accuracyVSAvoidreal-time processing capability
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent extracts only the necessary motion-related features (macro block types, motion vectors, bit numbers) from the compressed code stream without performing full decoding. This extraction approach obtains sufficient motion information while avoiding the heavy computational burden of complete picture decoding, thus achieving real-time processing capability while maintaining motion analysis accuracy

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent segments the video analysis process by working with individual macro blocks (16x16 pixel units) rather than entire pictures. By analyzing motion at the macro block level using compressed domain data, the system divides the computational task into smaller, more manageable units that can be processed in real-time while still providing accurate motion detection

Inventive Principle:
Principle #1Segmentation

2Ease of manufacture

If the same quantized coefficient is used for coding all video frames, then coding simplicity is maintained, but slowly moving objects that account for a smaller proportion of the total picture cannot be effectively detected as code stream size does not change significantly

Engineering Contradiction:
Improvecoding simplicityVSAvoidslowly moving object detection accuracy
Core Design Contradiction:
Ease of manufactureVSMeasurement precision

Solution Approach 1:

The patent applies local quality analysis by examining motion characteristics at the macro block level rather than using global code stream size analysis. By evaluating motion vectors and macro block types locally, the system can detect slowly moving objects even when they occupy small portions of the picture, while still using uniform quantization coefficients for coding simplicity

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent changes the analysis parameter from global code stream size to local macro block characteristics (motion vectors, macro block types, bit numbers). This parameter change enables detection of subtle motion in small regions without requiring different quantization strategies, maintaining coding simplicity while improving detection accuracy

Inventive Principle:
Principle #35Parameter changes

3Productivity

If code stream extraction method is used to reduce calculated quantity by extracting only reference frames, then computation load is reduced, but information may be missing as discarded frames may contain key motion information

Engineering Contradiction:
Improvecomputation efficiencyVSAvoidmotion information completeness
Core Design Contradiction:
ProductivityVSLoss of information

Solution Approach 1:

The patent implements a dynamic analysis approach where motion detection is performed on every frame's macro blocks using compressed domain data, rather than statically selecting only reference frames. By dynamically evaluating motion vectors and macro block types in each frame, the system maintains high computation efficiency while ensuring no motion information is lost, as each frame is independently analyzed for motion content

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP2782340B1Motion analysis method based on video compression code stream, code stream conversion method and apparatus thereof
Publication Date: 2017.11.01 HANGZHOU HIKVISION DIGITAL TECHNOLOGY CO LTD
  • EP2782340B1 patent drawingFigure 1
  • EP2782340B1 patent drawingFigure 2
  • EP2782340B1 patent drawingFigure 3~4

AI summary

The present invention relates to security video monitoring for and discloses a motion analysis method and a code stream conversion method based on video compression code stream and apparatus thereof. In the invention, a total number of bits or a residual bit number of a macro block in a video I frame s extracted from the video compression code stream; weighted average is performed to the total number of bits or a residual bit number of the macro block whose position corresponds with the current macro block in the latest M frames and a background value is obtained through calculating; a difference value between the total number of bits or the residual bit number of the current macro block and the background value is calculated, and if the difference value is greater than a predetermined first threshold, then a moving object is determined to be detected. Weighted average background modeling is performed to macro block information, such as the total numbers of bits or the residual bit number of the macro block corresponds with the current macro block in latest M frames. Threshold determination is performed to a background difference value matching result and motion analysis is performed. Coding picture of the macro block between the frames does not need to be calculated. Calculated quantity is small and the adaptability to the code stream and a scene is strong.