Motion Estimation Using Event Density Thresholds

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing motion estimation techniques for video streams are complex and costly due to the processing of entire frames, which includes redundant information, and struggle to efficiently capture scenes with varying brightness levels, leading to resource wastage and high computational complexity.

Innovation Solution

A method and system for motion estimation that segments frames into blocks, determines an event density factor for each block, compares it with a threshold value, estimates motion vectors based on the comparison, and processes blocks only where the event density factor exceeds the threshold, generating an event density map and block-level segmentation map to optimize block size for motion estimation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If entire frames are processed for motion estimation, then motion accuracy is maintained, but computational complexity and resource consumption increase significantly

Engineering Contradiction:
Improvemotion estimation accuracyVSAvoidcomputational complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides frames into multiple blocks and processes only those blocks where motion is detected, rather than processing entire frames. This segmentation approach maintains motion estimation accuracy in active regions while significantly reducing computational complexity in static regions.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts and processes only the relevant portions of frames (blocks with motion events) while discarding or skipping static blocks. This extraction principle eliminates redundant processing of unchanged regions, reducing resource consumption while preserving motion detection accuracy where needed.

Inventive Principle:
Principle #2Taking out (Extraction)

2Duration of action of stationary object

If fixed frame rate processing is used, then temporal coverage is uniform, but redundant information is transmitted wasting memory and energy

Engineering Contradiction:
Improvetemporal coverageVSAvoidenergy consumption
Core Design Contradiction:
Duration of action of stationary objectVSLoss of energy

Solution Approach 1:

The patent uses event-based sensors that operate asynchronously, triggering processing only periodically when changes occur rather than at fixed intervals. This allows the system to maintain temporal coverage of important events while reducing energy consumption by avoiding processing during static periods.

Inventive Principle:
Principle #19Periodic action

Solution Approach 2:

The patent discards redundant frame data where no motion occurs and only retains and processes blocks with detected changes. This selective retention strategy reduces memory usage and energy consumption while preserving all meaningful temporal information.

Inventive Principle:
Principle #34Discarding and recovering

3Measurement precision

If block matching is performed with all blocks of reference frame, then motion estimation precision improves, but processing resources required become prohibitively high

Engineering Contradiction:
Improvemotion estimation precisionVSAvoidprocessing resources
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts only the necessary blocks from reference frames for comparison - specifically those blocks corresponding to regions with detected motion events. This extraction eliminates the need to compare against all reference frame blocks, reducing processing resources while maintaining precision for active regions.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent performs partial block matching only where needed (in regions with motion events) rather than exhaustively matching all blocks. This partial action approach provides sufficient motion estimation precision for dynamic regions while avoiding the excessive resource consumption of complete frame-wide matching.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentUS11057641B1Systems and methods of motion estimation using monocular event-based sensor
Publication Date: 2021.07.06 SAMSUNG ELECTRONICS CO LTD
  • US11057641B1 patent drawing
  • US11057641B1 patent drawing
  • US11057641B1 patent drawing

AI summary

Provided is a method of motion estimation for processing a video stream comprising a plurality of frames, the method including segmenting at least one frame, from among the plurality of frames, into a plurality of blocks, determining an event density factor for each block included in a frame, wherein the event density factor of the block corresponds to a number of events accumulated in the block across frames in a predetermined time duration, comparing the determined event density factor with a threshold value, estimating a motion vector of the block based on the comparison, and processing the block in the video stream based on the estimated motion vector of the block.