Loading...
Please wait, while we are loading the content...
Similar Documents
Improvements to Seam Carving: A Content Aware Image Resizing Technique
| Content Provider | Semantic Scholar |
|---|---|
| Author | Kapadia, Dipti |
| Copyright Year | 2008 |
| Abstract | The recent developments in the field of display technologies have seen great diversity in display sizes. Same content is required to be displayed in different dimensions and aspect ratios for different devices. Standard scaling of an image to match the display size is not proving to be sufficient as it does not consider the image contents. Seam Carving is a content aware image resizing technique proposed by Ariel Shamir and Shai Avidan. It retargets images to a new size by gracefully carving out or inserting pixels in different parts of the image based on importance of pixels. In this thesis, the seam carving algorithm is analyzed for a variety of content, attention is then drawn to some of the image categories where it gives unacceptable results. A set of improvements by enhancing the so-called energy model which drives the seam carving algorithm are proposed. The new energy model reflects the visual saliency of image parts better than the original energy model. The improvements are evaluated in a subjective analysis. Also considering the popularity of wide screen displays, a retargeting algorithm for aspect ratio conversion using seam carving is proposed. Further on an evaluation of the performance of the proposed algorithm for aspect ratio change compared to currently known method of panoramic stretch is carried out. c © Koninklijke Philips Electronics N.V. 2008 iii TN-1 Philips Restricted Conclusions: Analysis of Seam carving, the proposed improvements to it and results of subjective evaluation leads to the following conclusions • Seam carving in general works best for images with multiple subjects separated by an uninteresting background. • The results of seam carving can be significantly improved by better modeling of visual saliency in the energy function. • Higher level cues like face detection greatly improves the results of retargeting. • Taking into account, the history of previously removed seams, continuous carving from the same region of image can be prevented leading to reduced artifacts. • Seam carving is a good candidate for retargeting for aspect ratio conversion as it considers the content while achieving target aspect ratio. iv c © Koninklijke Philips Electronics N.V. 2008 Philips Restricted TN-1 |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://alexandria.tue.nl/extra1/afstversl/wsk-i/kapadia2008.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
