Analysis of Image Matting Techniques Used for Generating Alpha-Matte
Sweta Singh, Dilip Kumar Sharma · 2023
The primary objective of automated matting techniques is to isolate the foreground object from its backdrop. To accurately fetch the foreground from the image, users must manually give some previous information to recognize foreground area, background area and unseen area in the form of trimap. Providing a well defined and accurate trimap by the user is a tedious job. Now a days various matting techniques try to focus on reducing the dependency on trimap, hence reducing the user burden. In this paper we have studied and analyzed various matting techniques. On the basis of required user intervention, we have broken down them into three categories: Manual, Semi-Automatic and Automatic Techniques. Finally the result of these techniques has been analyzed on the basis of their mean square error (MSE) and sum of absolute differences (SAD) parameters. Finally the paper concludes that Semi-Automatic and Automatic approaches reduce the user burden but they compromise with the result. Hence Manual trimap based methods outperforms Semi-Automatic and Automatic methods if trimap is specified carefully. TIMI-Net (a manual trimap based approach) is found to be best suited for the MSE as well as SAD parameter, showing least MSE value (0.0060) and least SAD value (29.08).