AI Image Processing Apparatus Noise Reduction Sharpness Preservation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image processing systems using neural networks struggle to balance noise reduction and preservation of image sharpness, often resulting in images with either excessive noise removal or loss of fine details.

Innovation Solution

An information processing apparatus that combines images processed using both AI and non-AI methods, applying weights based on sensitivity, brightness, and spatial frequency information to achieve optimal noise adjustment and sharpness preservation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Object-affected harmful factors

If noise reduction processing is performed using a neural network, then noise is removed from the input image, but fine details and sharpness are lost

Engineering Contradiction:

ImprovenoiseVSAvoidsharpness

Core Design Contradiction:

Object-affected harmful factorsVSManufacturing precision

Solution Approach 1:

The image is divided into multiple regions based on spatial frequency characteristics. High-frequency regions (edges, fine details) are processed differently from low-frequency regions (smooth areas). This segmentation allows selective noise reduction application, preserving sharpness in critical regions while reducing noise in others.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Different processing strategies are applied to different regions of the image. Regions with high spatial frequency content (containing fine details) receive lighter noise reduction to preserve sharpness, while regions with low spatial frequency content (smooth areas) receive stronger noise reduction. This local quality approach ensures that noise reduction does not uniformly degrade image sharpness.

Inventive Principle:

Principle #3Local quality

2Productivity

If traditional noise reduction methods are used, then processing speed is fast, but noise reduction effectiveness is insufficient

Engineering Contradiction:

Improveprocessing speedVSAvoidnoise

Core Design Contradiction:

ProductivityVSObject-affected harmful factors

Solution Approach 1:

The image is pre-processed by calculating spatial frequency information and dividing it into high-frequency and low-frequency regions before applying noise reduction. This preliminary segmentation enables the subsequent noise reduction algorithm to work more efficiently by focusing computational resources where they are most needed, rather than processing the entire image uniformly.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The noise reduction strength parameter is dynamically adjusted based on spatial frequency characteristics of different image regions. High-frequency regions receive lower noise reduction strength to preserve details, while low-frequency regions receive higher noise reduction strength for effective noise removal. This parameter adaptation allows the system to achieve good noise reduction effectiveness without excessive computational cost.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20230020328A1Information processing apparatus, imaging apparatus, information processing method, and program

Publication Date: 2023.01.19 FUJIFILM CORP

AI summary

There is provided an information processing apparatus including a processor and a memory connected to or built into the processor. The processor is configured to process a captured image by using an AI method that uses a neural network and perform composition processing of combining a first image obtained by processing the captured image by using the AI method and a second image obtained by processing the captured image without using the AI method.