Article Preview
TopImage retargeting is the modification in the image size, which is typically done in only a single direction in order to change the image aspect ratio. It emphasizes on not only change of width but also concentrate on its height i.e. it considers the image dimensions (Pritch et al., 2009). The traditional Image resizing methods involved uniform resizing, scaling and cropping the images and tries to preserve the salient objects (Maheswaran & Kumar, 2017). These methods may discard some important image information without worrying about the highlighted pixels. Due to this limitation, the methods are no longer used. To overcome the problems, Seam carving technique is developed which may emphasised on the high energy pixels. This technique deletes the low energy pixels seams and maintain the scene consistency. This can be enhanced (Wang et al., n.d.), by using a forward energy function which considers the energy inserted by deleting a seam. It was also extended to video retargeting.
Another retargeting algorithm is image warping. Image warping constructs a mesh for the image using a quadratic energy function. The image is deformed by using the vertices of this mesh. This preserves the salient contents of the image. In (Dong et al., 2009) a region based warping method is used to scale the objects uniformly and reduce the distortion of homogenous regions.
(Pritch et al., 2009) The logical approach of image retargeting is the Shift-map approach. In this approach, a shift-map is created for the image. The values of this map are adjusted to rearrange the image content. The efficient Importance filter is presented to calculate an integrated shift-map.
(Long et al., 2015) Semantic segmentation approach is used for the image retargeting to minimise the depth distortion. The method used convolutional network that train the pixels end to end and pixel by pixel to improve the performance in semantic segmentation.
Enormous efforts were made for retargeting the images in order to solve the depth distortion problem.