Image Processing With High-Frequency Detail Embedding

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image processing methods cause loss of high-frequency information during downsampling, leading to reduced image quality in upsampling, as they fail to account for content-dependent information in high-frequency components.

Innovation Solution

An image processing method that extracts content-dependent information from high-frequency components and writes it into low-frequency components, preserving details during downsampling and upsampling processes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If downsampling is performed to reduce image resolution, then storage and transmission costs are reduced, but high-frequency information is lost leading to reduced image quality after upsampling

Engineering Contradiction:

Improveimage data volumeVSAvoidhigh-frequency information

Core Design Contradiction:

Quantity of substanceVSLoss of information

Solution Approach 1:

The patent segments the image frequency components into high-frequency and low-frequency parts, then separately processes them. The high-frequency information is extracted and embedded into the low-frequency components during downsampling, allowing both components to be processed independently while preserving essential details.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent uses an intermediary embedding process where high-frequency information is mapped into the low-frequency image data. This intermediary step allows the high-frequency components to be carried within the low-frequency structure, enabling information preservation during compression.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Device complexity

If a same set of models is adopted for forward downsampling and backward upsampling, then processing simplicity is maintained, but image quality deteriorates due to loss of content-dependent high-frequency information

Engineering Contradiction:

Improveprocessing model complexityVSAvoidimage quality

Core Design Contradiction:

Device complexityVSManufacturing precision

Solution Approach 1:

The patent applies different processing strategies to different frequency components. Instead of using a uniform model for both downsampling and upsampling, it extracts content-dependent high-frequency information specifically and embeds it into low-frequency components, applying local quality enhancement where needed.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent performs preliminary extraction of high-frequency information during the downsampling process. By pre-extracting and embedding this information into the low-frequency components before storage or transmission, the system prepares the data in advance to maintain quality during subsequent upsampling operations.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20250225618A1Image processing method and apparatus, and storage medium and electronic device

Publication Date: 2025.07.10 DOUYIN VISION CO LTD

AI summary

An image processing method and apparatus, and a storage medium and an electronic device are provided. The method includes: acquiring an original image, and extracting high-frequency information and low-frequency information in the original image; extracting content dependent information in the high-frequency information, and writing the content dependent information into the low-frequency information, to obtain a low-resolution image corresponding to the original image.