Video Analysis Autogain Filter Using Texture Correlation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video surveillance systems face challenges in accurately distinguishing between foreground and background pixels due to automatic gain control effects, which can lead to incorrect object tracking and behavior analysis.

Innovation Solution

A method that extracts foreground patches from video frames using a background model, determines texture correlation scores, and removes pixels with high correlation scores to filter out false-positive foreground pixels caused by autogain, thereby refining the analysis.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Illumination intensity

If automatic gain control is used to improve dynamic range in low-light conditions, then image quality for human observers is improved, but the video analysis system's ability to distinguish foreground from background pixels deteriorates

Engineering Contradiction:
Improvedynamic rangeVSAvoidforeground-background pixel distinction
Core Design Contradiction:
Illumination intensityVSMeasurement precision

Solution Approach 1:

The patent segments the video frame into foreground and background regions using a background model, then applies texture analysis specifically to identify and remove false foreground pixels caused by autogain effects. This segmentation approach allows differential processing that preserves true foreground objects while eliminating autogain artifacts.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies local texture analysis to individual pixel regions rather than uniform processing across the entire image. By examining texture characteristics in local neighborhoods and comparing them against the background model, the system can selectively identify and remove only those pixels affected by autogain while preserving legitimate foreground content with different texture properties.

Inventive Principle:
Principle #3Local quality

2Measurement precision

If autogain is turned off to eliminate foreground-background distinction errors, then pixel distinction accuracy is improved, but image quality in low-light conditions deteriorates

Engineering Contradiction:
Improveforeground-background pixel distinctionVSAvoidimage quality in low-light
Core Design Contradiction:
Measurement precisionVSIllumination intensity

Solution Approach 1:

The patent converts the harmful autogain effect into a detectable pattern by analyzing texture characteristics. Instead of eliminating autogain entirely, the system uses its characteristic texture signature to identify and remove only the affected pixels, thereby maintaining the benefits of autogain for image quality while eliminating its harmful effects on analysis accuracy.

Inventive Principle:
Principle #22Blessing in disguise (Convert harm into benefit)

Solution Approach 2:

The patent changes the processing parameter from binary pixel intensity values to texture-based statistical measures. By transforming the data representation and analyzing higher-order texture properties rather than raw intensity values, the system can distinguish between legitimate foreground content and autogain artifacts even when both affect pixel intensity.

Inventive Principle:
Principle #35Parameter changes

3Productivity

If conventional background modeling is used without texture analysis, then processing speed is maintained, but accuracy in filtering false-positive foreground pixels deteriorates

Engineering Contradiction:
Improveprocessing speedVSAvoidfalse-positive pixel filtering accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies texture analysis selectively only to pixels identified as potential foreground candidates by the background model, rather than processing the entire image. This partial application of the more computationally intensive texture analysis maintains processing efficiency while achieving accurate false-positive filtering where it is most needed.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentUS10037599B2Automatic gain control filter in a video analysis system
Publication Date: 2018.07.31 INTELLECTIVE AI INC
  • US10037599B2 patent drawing
  • US10037599B2 patent drawing
  • US10037599B2 patent drawing

AI summary

Techniques are disclosed for analyzing a scene depicted in an input stream of video frames captured by a video camera. Bounding boxes are determined for a set foreground patches identified in a video frame. For each bounding box, the techniques include determining textures for first areas, each including a foreground pixel and surrounding pixels, and determining textures for second areas including pixels of the background model image corresponding to the pixels of the foreground areas. Further, for each foreground pixel in the bounding box area, a correlation score is determined based on the texture of the corresponding first area and second area. Pixels whose correlation scores exceed a threshold are removed from the foreground patch. The size of the bounding box may also be reduced to fit the modified foreground patch.