@conference {Mansfield2012, title = {Visibility maps for improving seam carving}, booktitle = {Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)}, volume = {6554 LNCS}, number = {PART 2}, year = {2012}, pages = {131{\textendash}144}, abstract = {In this paper, we present a new, improved seam carving algorithm. Seam carving efficiently removes pixels from an image to produce a retargeted image. It has proved popular with users and has been used as a component in many retargeting algorithms. We introduce the visibility map, a new framework for pixel removing image editing methods. This allows us to cast retargeting as a binary graph labelling problem. We derive a general algorithm which uses seam carving operations for efficient greedy optimization of a well defined energy, and compare this with forward energy seam carving and shift map image editing. We test this method with varying parameters on a large number of images, and present an improved seam carving algorithm which can demonstrably produce better results. We draw general conclusions about pixel removing methods for retargeting and motivate future directions of research. {\textcopyright} 2012 Springer-Verlag.}, isbn = {9783642357398}, issn = {03029743}, doi = {10.1007/978-3-642-35740-4_11}, url = {http://www.adobe.com/products/photoshop/photoshopextended/features/}, author = {Mansfield, Alex and Gehler, Peter and Van Gool, Luc and Carsten Rother} }