Announcement for Downloading full text filePlease respect the Copyright Act.
All digital full text dissertation and theses from this website are authorized the copyright owners. These copyrighted full-text dissertation and theses can be only used for academic, research and non-commercial purposes. Users of this website can search, read, and print for personal usage. In respect of the Copyright Act of the Republic of China, please do not reproduce, distribute, change, or edit the content of these dissertations and theses without any permission. Please do not create any work based upon a pre-existing work by reproduction, Adaptation, Distribution or other means.
URN etd-0826115-123414 Statistics This thesis had been viewed 1161 times. Download 67 times. Author Tung-Cheng Chen Author's Email Address No Public. Department Computer Science and Enginerring Year 2014 Semester 2 Degree Master Type of Document Master's Thesis Language Chinese&English Page Count 80 Title Image Auto-Cropping using Star-Light and Color Saturation Keyword saliency map saturation star-light star-light saturation saliency map Abstract Automatic image cropping can be used to segment the main theme of an image on different platforms such as mobile phones, cameras, and web pages. For the same input image the results are different in various systems with the same function. Some of the results may even be completely different. However, we should not say the result is wrong because it involves personal sense of aesthetics. Different methods may focus on different topics like
human faces or color contrast. Ma and Guo assessed regions based on entropy, size, and distance from the image center. Zhang et al. used face detection to find regions of interest.
Yen and Lin used training set before auto-cropping. Cheng used color contrast to generate a saliency map for cropping. Focused object is usually the saliency regions in an image. Performance of edge detection are often the measurement of saliency map. Therefore, we use L*a*b color space to
get color saturation value first and then use star-light mask to compare the pixel color saturation to generate an edge similar saliency map. Finally, calculate the pixel value for image cropping. Unlike ordinary well known Canny edges or Sobel edges, our method can generate better saliency map in which edges are kept while background edges are removed. Usability
and reliability are higher than face-based detection or color contrast based approaches. In addition, we propose two algorithms to automatically crop the main theme from saliency map. Our algorithm is both simple and easy to understand. It is an creative development approach compared with other auto-cropping method.
Advisor Committee Chen-Chiung Hsieh - advisor
Chueh-Wei Chang - co-chair
Mu-Chun Su - co-chair
Tsang-Long Pao - co-chair
Files Date of Defense 2015-07-28 Date of Submission 2015-08-26