|
|
July 2020 Issue Vol.10 No.7
Feature-based Thresholds on Alpha Matting for Images for Natural Image Dataset
https://ia601505.us.archive.org/14/items/vol10no701/vol10no701.pdf
Dr.T. Velumani1, Assistant Professor,
Department of Computer Science, Rathinam College of Arts and Science College (Autonomous), Coimbatore, TamilNadu, India
Abstract:
The aim of the research is to separate the foreground and background in natural images. The objects separation is performed by analysing the boundary area between foreground and background or the unknown. The
analysis of unknown area is used to determine the threshold value to separate definitive foreground and background in alpha matting. The process begins with defining a sub-image of the grayscale image dataset with
Region of Interest (ROI). Furthermore, the features of each sub-image consisting of contrast, correlation, energy and entropy are extracted using the Grey Level Co-Occurrence Matrix (GLCM) in angles of 0°, 45°, 90°, and 135°. Local
extraction results are averaged and normalized and then, it is treated as a threshold for alpha matting. The result is evaluated using Peak Signal Noise to Ratio (PSNR) and shows a significant increase in performance.
