Image Aspect Ratio Adaptation Using GAN Extension and ROI Cropping

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Images with mismatched aspect ratios between the image and display area result in undesirable white space, distortion, or loss of visual features when presented, as existing methods like color padding, blurring, or cutting off portions are inefficient.

Innovation Solution

A generative adversarial network (GAN) is used to extend images with additional features, followed by an augmented inference technique to reduce artifacts, and a machine learning model to identify and crop the region of interest, ensuring the image fits the display area without distortion or loss of significant features.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Area of stationary object

If traditional methods (color padding, blurring, cropping) are used to adapt images to different aspect ratios, then the display area can be filled, but visual quality deteriorates with white space, distortion, or loss of significant features

Engineering Contradiction:

Improvedisplay area coverageVSAvoidimage quality

Core Design Contradiction:

Area of stationary objectVSManufacturing precision

Solution Approach 1:

The system performs preliminary analysis of the input image to identify significant visual features (text, objects, regions of interest) before adapting the aspect ratio. This allows the system to plan the adaptation strategy in advance, ensuring that important features are preserved during the transformation process rather than discovering losses after adaptation.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system extends the image in the spatial dimension by generating additional pixel content beyond the original image boundaries. Using GANs, it creates realistic extensions that maintain consistency with the original image content, effectively increasing the image dimensions to accommodate different aspect ratios without distortion or cropping of significant features.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Area of stationary object

If the image aspect ratio is changed to fit the display area, then white space is eliminated, but significant visual features (text, objects) may be cut off or distorted

Engineering Contradiction:

Improvedisplay area utilizationVSAvoidvisual features

Core Design Contradiction:

Area of stationary objectVSLoss of information

Solution Approach 1:

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system introduces an intermediary processing stage between the original image and the final displayed image. This intermediary stage includes feature identification, extension generation using GANs, and intelligent cropping that prioritizes preserving significant visual features. The intermediary processes ensure that no important information is lost during aspect ratio adaptation.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Manufacturing precision

If image extension is performed to maintain all visual features, then image quality is preserved, but the process becomes more complex requiring machine learning models

Engineering Contradiction:

Improveimage qualityVSAvoidprocessing system complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The system replaces traditional mechanical image processing methods (resizing, cropping, padding) with machine learning-based approaches. Specifically, it uses Generative Adversarial Networks (GANs) to intelligently extend image content and preserve visual features, substituting complex computational processes for simple geometric transformations.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system employs self-service mechanisms where the machine learning models automatically analyze image content, identify significant features, and perform appropriate extensions without manual intervention. The GANs self-adjust to generate realistic extensions based on the input image characteristics, reducing the need for manual parameter tuning or complex configuration.

Inventive Principle:

Principle #25Self-service

4Manufacturing precision

If machine learning models are used to extend images, then image quality and realism are improved, but computational time and processing resources increase

Engineering Contradiction:

Improveimage realismVSAvoidprocessing time

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The system applies partial action by focusing computational resources only on extending the necessary portions of the image rather than processing the entire image uniformly. It identifies regions that need extension and applies GANs selectively to those areas, reducing overall processing time while maintaining image realism where needed.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20260017756A1Flexible image aspect ratio using machine learning

Publication Date: 2026.01.15 GOOGLE LLC

AI summary

To adjust an aspect ratio of an image to match the aspect ratio of a display area for presenting the image, a computing device receives an image having a first aspect ratio, and obtains a second aspect ratio for a display area of a display in which to present the image, where the second aspect ratio is different from the first aspect ratio. The computing device extends the image to include one or more additional features which were not included in the image. Additionally, the computing device automatically crops the extended image around an identified region of interest by selecting a portion of the extended image that has an aspect ratio which matches the second aspect ratio of the display area, and provides the cropped image for presentation within the display area of the display.