Dynamic Image Downsampling Using Neural Network Models

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image downsampling methods, such as image interpolation, result in poor image quality and lack flexibility in downsampling to any desired size.

Innovation Solution

An image processing method that determines a target downsampling rate and network model based on the original and target image sizes, and then downsamples a second image using this model to achieve the target size while enhancing image quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Area of stationary object

If image interpolation is used for downsampling, then the image size is reduced, but the image quality deteriorates

Engineering Contradiction:
Improveimage sizeVSAvoidimage quality
Core Design Contradiction:
Area of stationary objectVSManufacturing precision

Solution Approach 1:

The patent changes the parameter of downsampling rate to be variable rather than fixed. By allowing the downsampling rate to be dynamically adjusted based on the relationship between original and target image sizes, the system can achieve accurate downsampling to any desired size while maintaining image quality through optimized processing parameters.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent replaces the traditional mechanical image interpolation method with a neural network-based downsampling approach. The neural network model learns optimal downsampling transformations during training, substituting the conventional interpolation mechanics with a data-driven model that preserves image quality while achieving size reduction.

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

2Device complexity

If fixed downsampling rate is used, then the processing is simple, but the flexibility to downsample to any size is limited

Engineering Contradiction:
Improveprocessing complexityVSAvoiddownsampling flexibility
Core Design Contradiction:
Device complexityVSAdaptability or versatility

Solution Approach 1:

The patent introduces dynamics by making the downsampling rate variable and adaptive. Instead of using a fixed rate, the system dynamically calculates the appropriate downsampling rate based on the target image size requirements, allowing flexible downsampling to any desired size while managing processing complexity through automated rate determination.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent achieves universality by enabling the same downsampling system to handle multiple target sizes through a single neural network model. The model is trained to perform downsampling at various rates, making it multi-functional and adaptable to different image size requirements without requiring separate processing pipelines for each size.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentUS20250173824A1Image processing method, apparatus, device and storage medium
Publication Date: 2025.05.29 DOUYIN VISION CO LTD
  • US20250173824A1 patent drawing
  • US20250173824A1 patent drawing
  • US20250173824A1 patent drawing

AI summary

Embodiments of the present disclosure provide an image processing method and apparatus, a device and a storage medium. The method comprises: determining, based on an original size and a processed target size corresponding to a first image to be processed, a target downsampling rate corresponding to the first image; determining a target downsampling network model corresponding to the first image based on the target downsampling rate, at least one pre-trained and obtained downsampling network model and a preset downsampling rate corresponding to the downsampling network model; determining a second image satisfying a preset downsampling condition based on a preset downsampling rate corresponding to the target downsampling network model, the target downsampling rate and the first image; downsampling the second image based on the target downsampling network model to obtain a target image having the target size.