Multi-Resolution Image Noise Reduction via Motion Compensation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image processing methods are inefficient in reducing noise in video signals, particularly when dealing with motion between frames, as they require significant computational resources and memory access, which can lead to increased power consumption and reduced performance.

Innovation Solution

An image processing device and method that decomposes input images into multiple resolution levels, detects motion vectors between lower-resolution base images and preceding frame images, performs motion compensation, and combines the results with intra-frame noise reduction to generate noise-reduced images, thereby reducing the load on frame memory and improving processing efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If motion compensation is performed on high-resolution images using full-frame memory, then noise reduction quality is improved, but memory capacity requirements and power consumption increase significantly

Engineering Contradiction:

Improvenoise reduction qualityVSAvoidframe memory capacity

Core Design Contradiction:

Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent divides the image processing into multiple resolution levels (full-resolution and reduced-resolution). Motion compensation is performed on reduced-resolution images using a portion of frame memory, while full-resolution images undergo different processing. This segmentation allows noise reduction to be achieved on manageable memory subsets rather than requiring entire high-resolution frame buffers.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces a resolution dimension to the processing hierarchy. By creating and processing reduced-resolution versions of images, the system operates in an additional resolution dimension, allowing motion compensation to be performed on smaller data representations that require less memory capacity while still contributing to overall noise reduction.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Measurement precision

If motion vector detection and motion compensation are performed on full-resolution images, then processing accuracy is improved, but computational complexity and processing time increase

Engineering Contradiction:

Improvemotion detection accuracyVSAvoidprocessing complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the computational workload by resolution level. Reduced-resolution images are used for motion vector detection and motion compensation operations that are computationally intensive, while full-resolution images are processed separately. This segmentation reduces the computational complexity of each processing stage while maintaining overall system accuracy.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies motion compensation partially - performing it on reduced-resolution images rather than all images at full resolution. This partial action approach reduces computational complexity significantly while still achieving noise reduction benefits, accepting that not every image undergoes the full motion compensation process.

Inventive Principle:

Principle #16Partial or excessive action

3Measurement precision

If cyclic noise reduction is performed on all frames at full resolution, then noise reduction effectiveness is improved, but power consumption and processing load increase

Engineering Contradiction:

Improvenoise reduction effectivenessVSAvoidpower consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The patent segments the noise reduction process by applying different processing strategies to different resolution levels. Cyclic noise reduction is performed on reduced-resolution images that require less computational power and memory access, thereby reducing overall power consumption while maintaining effective noise reduction through the multi-resolution approach.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies cyclic noise reduction partially - primarily on reduced-resolution images rather than all full-resolution frames. This partial application reduces the power consumption and processing load associated with cyclic noise reduction while still achieving effective noise reduction when results are combined across resolution levels.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS10825145B2Image processing device, image processing method, and non-transitory computer readable medium storing image processing program

Publication Date: 2020.11.03 OLYMPUS CORPORATION(JP)

AI summary

An image processing device includes: a frame memory that saves a cyclic image of an immediately preceding frame as a reference image; and a processor comprising hardware. The processor is configured to: decompose an input image into a plurality of resolution images; set, as a base image, an image having a lower resolution than the input image among the plurality of resolution images and that detects a motion vector between the base image and the reference image; subject the reference image to motion compensation on the basis of the motion vector to generate a motion-compensated image; combine the base image with the motion-compensated image to generate a noise-reduced base image that serves as a new cyclic image; and combine the noise-reduced base image with the resolution image or resolution images other than the base image among the plurality of resolution images.