Image Noise Filtering via Laplacian and Gaussian Pyramid Compositing

Solution Overview

Problem

Conventional image processing methods either fail to effectively suppress noise due to compositing unevenness when using non-overlapping frequency bands or suffer from reduced detection accuracy and image quality when using overlapping bands.

Innovation Solution

An image processing apparatus and method that separates images into luminance and color frequency bands using both Laplacian and Gaussian pyramids, applying noise suppression to each band and recompositing them with a calculated ratio based on edge signals to alleviate compositing unevenness and enhance noise suppression.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Stability of the object's composition

If an image is separated into a plurality of bands with no overlapping frequency bands using the Laplacian pyramid, then compositing unevenness is reduced, but noise suppression effect is difficult to improve

Engineering Contradiction:

Improvecompositing unevennessVSAvoidnoise suppression effect

Core Design Contradiction:

Stability of the object's compositionVSReliability

Solution Approach 1:

The image is segmented into multiple frequency bands using both Laplacian pyramid (for luminance) and Gaussian pyramid (for color), allowing independent noise suppression processing for each band while maintaining the ability to control compositing through edge signal analysis

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Different pyramid methods are applied to different signal types: Laplacian pyramid for luminance signals where compositing stability is crucial, and Gaussian pyramid for color signals where noise suppression effectiveness is more important, optimizing each channel's processing characteristics

Inventive Principle:

Principle #3Local quality

2Reliability

If an input image is separated into a plurality of images with overlapping frequency bands using the Gaussian pyramid, then noise suppression effect is improved, but compositing unevenness occurs due to reduced detection accuracy for edge signals

Engineering Contradiction:

Improvenoise suppression effectVSAvoidcompositing unevenness

Core Design Contradiction:

ReliabilityVSStability of the object's composition

Solution Approach 1:

Different pyramid methods are applied to different signal types: Laplacian pyramid for luminance signals where compositing stability is crucial, and Gaussian pyramid for color signals where noise suppression effectiveness is more important, optimizing each channel's processing characteristics

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

Edge signals are extracted and analyzed to provide feedback for calculating compositing ratios, allowing the system to dynamically adjust the blending of overlapping frequency bands to minimize compositing unevenness while maintaining noise suppression benefits

Inventive Principle:

Principle #23Feedback

3Reliability

If a compositing ratio is calculated based on extracted edge signals, then control of noise amount is improved, but image quality deteriorates due to reduced detection accuracy for edge signals

Engineering Contradiction:

Improvenoise controlVSAvoidimage quality

Core Design Contradiction:

ReliabilityVSManufacturing precision

Solution Approach 1:

Different pyramid methods are applied to different signal types: Laplacian pyramid for luminance signals where compositing stability is crucial, and Gaussian pyramid for color signals where noise suppression effectiveness is more important, optimizing each channel's processing characteristics

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

Edge signals serve as an intermediary element that enables intelligent compositing ratio calculation, allowing the system to automatically adjust blending parameters based on local image characteristics without requiring manual intervention or sacrificing overall image quality

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS9317901B2Image processing apparatus and image processing method for noise filtering

Publication Date: 2016.04.19 CANON KK

AI summary

An input image is divided into images composed of luminance signals of a plurality of bands that make up the Laplacian pyramid, noise suppression is applied to each divided image, and then the divided images are composited by addition. The input image is also divided into images composed of color signals of a plurality of bands that make up the Gaussian pyramid, noise suppression is applied to each divided image, and the divided images are composited at an image-based ratio. By thus compositing the luminance signals and color signals, excellent noise suppression can be realized while alleviating deterioration in the image quality during the compositing.