CASAIR: Content and Shape-Aware Image Retargeting and Its Applications

Area of Science:

• Computer Vision
• Image Processing
• Computer Graphics

Background:

• Traditional image retargeting often struggles with non-rectangular shapes.
• Existing methods may compromise image content or shape fidelity.

Purpose of the Study:

• To propose a novel image retargeting algorithm (CASAIR) for non-rectangular shapes.
• To preserve image content and adapt to target shapes simultaneously.
• To introduce a new class of shapes, bhv-convex shapes, achievable by the algorithm.

Main Methods:

• CASAIR employs seam segment carving to remove low-cost segments.
• A cost function integrates image content and target shape information.
• The algorithm is demonstrated with a smart camera-projector system for online retargeting.

Main Results:

• CASAIR successfully retargets images to a wide range of non-rectangular shapes.
• The algorithm achieves dual objectives of content preservation and domain transformation.
• Bhv-convex shapes are identified as the achievable target shapes.

Conclusions:

• CASAIR offers an efficient and flexible platform for image retargeting.
• The algorithm enhances image fidelity in ubiquitous display systems.
• CASAIR broadens the application scope of image retargeting significantly.