Seam Carving Image Resizing with Saliency Preservation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for resizing content items, such as images and videos, often distort salient content due to inadequate preservation of important pixels during resizing processes, particularly in mixed content environments like web pages that include text, faces, and logos.

Innovation Solution

The implementation of seam carving technology that utilizes saliency maps to select and manipulate seams based on pixel importance, ensuring that primary content is preserved by duplicating or removing connected pixels with low saliency scores, while adhering to seam cost thresholds and iterative resizing processes to maintain image quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If traditional resizing methods are used to change image dimensions, then the image can be resized to fit different slot dimensions, but the salient content (text, faces, logos) becomes distorted

Engineering Contradiction:
Improveimage dimension adaptabilityVSAvoidcontent preservation accuracy
Core Design Contradiction:
Adaptability or versatilityVSManufacturing precision

Solution Approach 1:

The image is divided into multiple seams (vertical or horizontal bands of pixels) that can be selectively removed or duplicated. Each seam represents a segment of the image that can be independently manipulated based on its saliency score, allowing precise control over which parts of the image are modified during resizing.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different regions of the image are treated differently based on their local importance. Saliency maps identify regions with text, faces, and logos as high-importance areas that should be preserved, while other regions are marked as lower importance and can be modified. This local differentiation allows the resizing algorithm to protect critical content while still achieving the desired dimension changes.

Inventive Principle:
Principle #3Local quality

2Manufacturing precision

If seam carving is applied to preserve salient content, then content integrity is maintained, but the processing complexity and time increase

Engineering Contradiction:
Improvecontent preservation accuracyVSAvoidresizing process complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

Before performing the actual seam removal or duplication operations, the algorithm first computes a saliency map that identifies important regions in the image. This preliminary analysis phase prepares the necessary information (saliency scores for each pixel and seam) in advance, allowing the subsequent resizing operations to proceed efficiently with clear guidance on which seams to modify.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The saliency map computation and seam selection process is fully automated, requiring no manual intervention. The algorithm automatically identifies salient regions, calculates seam costs, selects optimal seams for removal or duplication, and performs the resizing operations all through self-contained computational processes, reducing the need for external control or complex user input.

Inventive Principle:
Principle #25Self-service

3Length of moving object

If multiple seams are removed or duplicated to achieve significant resizing, then the dimension change is substantial, but the risk of distorting important content increases

Engineering Contradiction:
Improveimage dimension changeVSAvoidcontent distortion risk
Core Design Contradiction:
Length of moving objectVSReliability

Solution Approach 1:

The algorithm uses the saliency map as continuous feedback to guide seam selection throughout the resizing process. By constantly referencing the pre-computed saliency scores, the algorithm ensures that seams with low saliency (less important regions) are selected for removal or duplication, while seams passing through high-saliency regions (text, faces, logos) are avoided. This feedback mechanism maintains content integrity even when significant dimension changes are required.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The resizing process dynamically adjusts the number and position of seams to be removed or duplicated based on the desired output dimensions. The algorithm calculates the required dimension change, then selects the appropriate number of low-saliency seams to achieve that change, adapting the resizing strategy to match the specific requirements while maintaining content preservation.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP2948919B1Systems and methods for resizing an image
Publication Date: 2019.01.16 GOOGLE LLC
  • EP2948919B1 patent drawingFigure 1
  • EP2948919B1 patent drawingFigure 2
  • EP2948919B1 patent drawingFigure 3A~3B

AI summary

In some instances, an image may have dimensions that do not correspond to a slot to display the image. For example, an image content item may have dimensions that do not correspond to a content item slot. The image may be resized using seam carving to add or remove pixels of the image. A saliency map for the image may be used having saliency scores for each pixel of the image. Evaluation metrics may be used before, during, and after, seam carving to determine whether salient content is affected by the seam carving. In some instances, a seam cost threshold value may be used for adaptive step size during the seam carving. The resized image may then be outputted, such as for an image content item to be served with a resource.