Guided Filter Detail Enhancement for Video Signals

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video processing methods struggle to enhance fine details in video signals without causing overshoot or ringing on edges, and fail to efficiently enhance details in dark or smooth regions, leading to inadequate detail enhancement and potential temporal flicker.

Innovation Solution

A system utilizing a guided image filter to generate a detail signal by taking a difference between the input video signal and a linear transformation of a guidance signal, with a noise reduction step and a gain function based on variance, to produce a detail-enhanced video signal that effectively adds details to the input signal while preserving edges and reducing noise.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If un-sharp masking is used for fine detail enhancement, then small details are enhanced, but overshoot or ringing occurs on edges

Engineering Contradiction:
Improvedetail enhancement qualityVSAvoidovershoot or ringing on edges
Core Design Contradiction:
Manufacturing precisionVSObject-generated harmful factors

Solution Approach 1:

The patent introduces a guided filter as an intermediary component between the un-sharp masking operation and the final output. This guided filter processes the enhanced signal to remove overshoot and ringing artifacts while preserving the enhanced details, effectively mediating between the conflicting requirements of detail enhancement and edge quality

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces the traditional un-sharp masking mechanical filtering approach with a guided filter-based processing system. This substitution allows for more controlled and adaptive filtering that can enhance details without producing the harmful overshoot and ringing effects characteristic of conventional un-sharp masking

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Object-affected harmful factors

If traditional filtering is used to remove noise, then noise is reduced, but fine details in dark or smooth regions are lost

Engineering Contradiction:
Improvenoise levelVSAvoidfine detail visibility in dark regions
Core Design Contradiction:
Object-affected harmful factorsVSManufacturing precision

Solution Approach 1:

The guided filter applies different filtering strengths to different regions of the image based on local characteristics. In dark or smooth regions where details are vulnerable, the filter adapts to preserve fine structures while still reducing noise, rather than applying uniform filtering across the entire image

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The filtering operation is made dynamic and adaptive rather than static. The guided filter adjusts its behavior based on local image content, allowing it to be more aggressive in noisy regions while being gentler in dark or smooth regions to preserve fine details that would otherwise be lost

Inventive Principle:
Principle #15Dynamics

3Manufacturing precision

If detail enhancement is applied to improve transitions, then small details are enhanced, but temporal flicker occurs

Engineering Contradiction:
Improvedetail enhancement qualityVSAvoidtemporal stability
Core Design Contradiction:
Manufacturing precisionVSStability of the object's composition

Solution Approach 1:

The patent incorporates feedback mechanisms that monitor the enhanced output and adjust the processing to minimize temporal variations. By using the guided filter to stabilize the enhanced signal, the system can maintain detail enhancement quality while reducing the temporal flicker that would otherwise occur

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The guided filter processing acts as a cushioning layer that prevents the harmful temporal flicker effects from manifesting in the final output. By processing the enhanced signal through this stabilizing filter before output, the system cushions against the temporal instability that would otherwise be introduced by detail enhancement

Inventive Principle:
Principle #11Beforehand cushioning (Prior cushioning)

Data Source

PatentEP2976877B1Guided filter-based detail enhancement
Publication Date: 2020.01.08 SYNAPTICS INC
  • EP2976877B1 patent drawingFigure 1~2
  • EP2976877B1 patent drawingFigure 3A~3B
  • EP2976877B1 patent drawingFigure 3C

AI summary

Systems and methods for generating a detail-enhanced video signal are provided. In a method for generating the detail-enhanced video signal, an input video signal is received. A first signal that is a linear transformation of the input video signal is generated, A detail signal is generated by determining a difference between the input video signal and the first signal, where the detail signal includes information that is added to the input video signal to generate the detail-enhanced video signal. A filtered detail signal is generated by removing noise from the detail signal. The filtered detail signal is multiplied by a gain function to generate a second signal. The gain function is based on an amount of variance in the input video signal. The input video signal and the second signal are combined to generate the detail-enhanced video signal.