Image retargeting with multifocus fisheye transformation

Seam warping: a new approach for image retargeting for small displays

“…These methods are called content-aware image retargeting (Zhang et al 2013). Different types of content-aware image retargeting techniques have been developed in recent years, such as content-aware cropping, Ciocca et al (2007) and Nishiyama et al (2009), segmentation-based approach (Setlur et al 2005), seam carving (Avidan and Shamir 2007;Rubinstein and Shamir 2008), patch-based approach (Barnes et al 2009;Lin et al 2013), warping-based approach (Zhang et al 2013;Wang et al 2011), and multi-operator approach (Rubinstein and Shamir 2009;Luo et al 2012). All of these techniques have their own advantages and disadvantages.…”

“…Compared with other retargeting methods, the warpingbased approach leads to a smoother result without loss of image information, that is, the target image contains the "focus+context" of an image. Existing warping-based methods can be categorized into three classes according to the type of objects a warping function targets to: (1) the regions of an image (Zhang et al 2013;Wolf et al 2007;Liu and Gleicher 2005;Zhang et al 2010), (2) the pixels of an image (Kim et al 2009;Ren et al 2009;Brand 2009), and (3) the mesh of an image (Wang et al 2008;Zhang et al 2009;Guo et al 2009). All of these methods are faced with a common problem, namely, how to find out the ROI in the source image.…”

“…All of these techniques have their own advantages and disadvantages. Detailed analysis has been conducted in our previous paper (Zhang et al 2013), so we don't poll-parrot in this paper. Considering the focus on this research, our review here focuses on two of them: seam caving and warping.…”

“…This problem is especially serious when the warping function is based on region transformation. Fisheye-based warping method (Zhang et al 2013;Liu and Gleicher 2005;Zhang et al 2010) which belongs to the first kind of the above circumstance is a case in point.…”

“…However, existing methods often have limitations (Zhang et al 2013). Scaling methods may shrink the important areas of the image and make the image unrecognizable when the resolution of the original image is significantly higher than that of the target screen.…”