| Summary: | In Singapore, a common sight is one holding his or her mobile phone taking photographs. These users take a large number of digital photographs and sometimes, they wish to edit it. For example, one may wish to cut out his or her picture (object image) from a group photo and paste it onto another image (scene image) that shows an Egyptian pyramid. The edited image will show the user standing in front of the pyramid regardless of whether the user has been to Egypt before. In this project, there are two approaches to segment the object image from a test image, automated and manual. The automated approach is to automatically detect the edges of object image and segment it from the test image. The manual approach is to manually define the edges of the object image so that it can be segmented from the test image. The automated approach includes two methods, MATLAB Canny Edge Detection and MATLAB active contour whereas the manual approach uses MATLAB specifying region-of-interest (ROI) method. It was tested and verified that the manual approach performs better as the edges of object image can be clearly defined and segmented. By using MATLAB indexing method, the object image can be superimposed onto any part of the scene image. Thereafter, image processing techniques such as weighted pixel averaging and thresholding are used to blend the object and scene images so that the object image looks natural in the scene image. Some of the important concepts that will be reviewed include Canny Edge Detection algorithm and superimposition by indexing. Testing and implementation have been done and pictorial views of the results are also included.
|
|---|