Refooormat - Mobile Image Retargeting

This is my Bachelor thesis on content-aware image retargeting. It is a universal iOS app that allows real-time refinement of the retargeted image.

Visit the website of the project for more information: http://refooorm.at

The app is available in the iOS app store: https://itunes.apple.com/app/refooormat/id650410245

This code is released under GPLv3.

Mobile Image Retargeting
Bachelor Thesis
Daniel Graf
grafdan@ethz.ch

Technical ReadMe

The Xcode project runs out of the box. Just press run to execute it in the iOS simulator or choose your connected device (needs code signing).

Project Structure

The classes are grouped into:

• CVXGen Solver
• Model (keep state of projects, retargeting operator, settings...)
• Controller (connect UI with OS and state)
• View (everything that draws onto the screen) -- Toolbars (Editor, Library) -- Popovers (Settings, Ratio Picker) -- RetargetingViews (OpenGL code)
• Helper (small extensions)
• Numerics (compressed column storage matrix library)

Key Parts of the Implementation

• retargeting operator (prepare and call solver for convex QP) --> RetargetingSolver solveASAPWithTask:
• cropping --> RetargetingSolver solveWithCropping:
• threshold range interpolation --> RetargetingSolver croppingRampWidth:withTask:
• automatic saliency --> RetargetingState automaticSaliency: calculateGradient: / faceDetection:
• painting and saliency matrix update RetargetingState paintAtPosition:
• connect everything together: RetargetingViewController
• OpenGL calls to render the image: e.g. in CombineSaliencyView drawRect:
• fixed-point motion stabilization CombinedContainerView paintImage: CombinedVew layoutSubviews: