Seam carving

Content-based image distortion (also known as an intelligent, picture - or content-based photo or image distortion, image scaling, known in English as Seam carving, retargeting, Content Aware Image Resizing - CAIR, Content Aware Scaling, Liquid Resizing or Liquid rescaling ) is a patented method for distortion the aspect ratio of images with preservation of proportions and shapes of relevant image content and thus with minimal perceived distortion. It was originally introduced by Shai Avidan from Mitsubishi Electric Research Labs ( MERL ) and Ariel Shamir ( Interdisciplinary Center and MERL ) at SIGGRAPH 2007.

The algorithm creates a Entbehrlichkeitswertung for all areas / pixels of the image and studied using this then paths unimportant as possible pixels that extend transversely to upset or stretching direction and adds to these joints (English seams, → seam carving ) pixel or removed any. Different variants of the basic algorithm differ in the rule in the strategies for creating the Entbehrlichkeitswertung, which is particularly challenging for moving images. In the original, the simplest case, this is done only based on the size of Tonwertunterschieden between adjacent pixels (energy). For this weighting itself offers the possibility of human intervention in the automatic transmission and is available in many deployments. This also entire objects can for example be marked for complete removal.

Adobe Systems acquired by the MERL a simple license to use the method and implemented it in Photoshop CS4 (Select the " Content Aware Scaling" ). Other popular graphics software like GIMP, digiKam and ImageMagick provide implementations of the technology.

Improvements and enhancements

  • Better energy function and applied to moving images ( video)
  • Combination with cropping and scaling
200935
de