ABSTRACT
Image content is nowadays visualized on a variety of displays, from the small screen of a cellphone to a large home TV. Many websites display images whose size is dynamically adapted to the browser window size and the surrounding text. All these display options require different formats and resolutions, and the images need to be scaled to fit the available display space. If the aspect ratio of a display is the same as the image, uniform rescaling can be applied, but otherwise, an image needs to be retargeted to the display aspect ratio. The two most common strategies are cropping, which uniformly scales the image to fit the entire screen and discards the parts that do not fit, and letterboxing, which adds black bands around the image. To prevent losing parts of the image and at the same time avoid wasting screen space, content-aware retargeting techniques were developed. These methods selectively deform the input image into the target dimensions according to a saliency map, preserving the shape of important image components while distorting unimportant background content (see Figure 5.1). A few general methodologies for retargeting were proposed in recent years, such as discrete carving, continuous warping, and hybrid
approaches [246] and they are even available in modern commercial image editing software [5].
