Please use this identifier to cite or link to this item:
Title: Image abstraction
Authors: Pay, Chin Yeen.
Keywords: DRNTU::Engineering::Computer science and engineering::Computer applications::Arts and humanities
DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
Issue Date: 2012
Abstract: This report will present the experimental results of a non-photo realistic rendering (NPR) technique that will transform an image into a watercolor abstraction. NPR enable an artist to convert the image into many expressive styles. Winnemoller, Olsen and Gooch (2006) [13] propose several steps to transform the image. This is done by using a bilateral filter to smoothen the non-edges, using a difference of Gaussian (DOG) to highlight the edges and lastly, quantizing the luminance value. This technique has several weaknesses and strengths that will be explored in the paper. It will present the results of the filter on various sample images and finally a video. This technique is able to work fairly well for the images, and the images are turned into the expected cartoonish images. However, the DOG filter appears to have difficulties detecting edges within regions that have very dark colors. From the experimental results, we will also not recommend the use of this filter on videos running on real time as the processing time of each frame can be unpredictable. The results of the experiment will be evaluated based on how effective the technique is able to denoise the image, identify the edges and lastly quantize the colors of the image, and how the different parameters and features of an image affects the algorithm. We recommend the use of this technique on images with very distinct color regions that are bright against the edges.
Rights: Nanyang Technological University
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:SCSE Student Reports (FYP/IA/PA/PI)

Files in This Item:
File Description SizeFormat 
  Restricted Access
2.66 MBAdobe PDFView/Open

Page view(s) 50

Updated on Nov 23, 2020

Download(s) 50

Updated on Nov 23, 2020

Google ScholarTM


Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.