GPU Bilateral Filter for Real-Time Video Noise Cancellation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing techniques for real-time processing of video feeds from multiple cameras in airport surveillance are hindered by the computational expense of bilateral filters, making them unsuitable for real-time applications.

Innovation Solution

Implementing a bilateral filter within a GPU shader, utilizing a programmable shader on a graphics processing unit (GPU) to efficiently process large video images in real-time, with precomputed kernel coefficients optimizing computation time.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If bilateral filters are applied to video feeds for noise cancellation, then image clarity and noise reduction are improved, but computational expense increases making real-time processing impossible

Engineering Contradiction:
Improveimage clarityVSAvoidreal-time processing speed
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The patent segments the video processing task by separating noise reduction from edge preservation operations. The bilateral filter is decomposed into independent spatial and range components that can be processed separately and then combined, reducing overall computational complexity while maintaining real-time performance capability

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the parameters of the bilateral filter by precomputing and storing filter kernels in a lookup table. By transforming the filter into a form where kernels depend only on gradient magnitude rather than both spatial and range distances, the computational parameters are simplified, enabling real-time processing without sacrificing image clarity

Inventive Principle:
Principle #35Parameter changes

2Object-affected harmful factors

If bilateral filters are used for noise reduction, then noise cancellation is improved, but device complexity and computational resources required increase

Engineering Contradiction:
Improvenoise reductionVSAvoidcomputational complexity
Core Design Contradiction:
Object-affected harmful factorsVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by precomputing bilateral filter kernels and storing them in a lookup table before video processing begins. This precomputation eliminates the need for complex real-time calculations during video frame processing, significantly reducing device complexity while maintaining effective noise reduction capability

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent uses copying by creating a lookup table that stores precomputed filter kernels. Instead of recalculating complex bilateral filter weights for each pixel and frame, the system copies and applies pre-stored kernel values, dramatically reducing computational requirements while preserving noise reduction effectiveness

Inventive Principle:
Principle #26Copying

Data Source

PatentEP3384673B1Noise-cancelling filter for video images
Publication Date: 2022.01.26 SEARIDGE TECH
  • EP3384673B1 patent drawingFigure 1
  • EP3384673B1 patent drawingFigure 2
  • EP3384673B1 patent drawingFigure 3

AI summary

A video processing system and method apply a bilateral filter to images of a video stream in real time. The bilateral filter is executed and applied using a graphics processing unit (GPU) controlled by a processor. The bilateral filter may be encoded in a shader operated by the GPU. The GPU or processor may be configured to compress one or more video images of the video stream. Blurring or smoothing of the video images by the shader-implemented bilateral filter may reduce image noise thereby increasing a compression performance. The bilateral filter may be applied exclusively to a background of the video images which are substantially free of sharp edges. The video stream may be received from cameras covering an area of an airport, which may be an airport apron.