New Controls for Combining Images in Correspondence.

Area of Science:

• Computer Vision
• Image Processing
• Computer Graphics

Background:

• Existing image interpolation methods often use simple geometric and color blending functions.
• Defining correspondence maps for structurally similar elements is a key challenge in morphing.

Purpose of the Study:

• To explore new controls for combining images using correspondence maps.
• To develop an intuitive framework for merging shape and color information from source images.
• To enable interactive synthesis of hybrid images and video morphing.

Main Methods:

• Applied edge-aware decomposition techniques to correspondence maps, not just image content.
• Developed a low-dimensional parameter space for merging shape and color at multiple frequencies.
• Created a gallery-based user interface for interactive exploration and synthesis.

Main Results:

• Established an intuitive parameter space for merging image attributes.
• Demonstrated compelling effects through extrapolation of shape parameters.
• Extended the framework successfully to video interpolation.

Conclusions:

• The proposed framework offers enhanced control over image interpolation and morphing.
• Edge-aware decomposition of correspondence maps is effective for richer image blending.
• The method facilitates intuitive creation of hybrid imagery and video effects.