Password authentication is failing as an authentication since it increases the user burden to remember the passwords. Graphical authentication is proposed as a alternative for textual passwords since it may be easy for users to remember. In this paper we propose a new image region selection based graphical password scheme. We are going to present a new technique for authentication which is based on the tracking of mouse motions on an image called mouse gestures for selecting regions in the image.A set of gestures may be stored in a database called gesture classes for each user. Users are allowed to select a set of random images and a gesture for each image. Some tolerance level is also given for each gesture. When logging in if the user draws the correct gesture using mouse the user will be treated as an authenticated user.Mouse gestures are captured through bounding box and corner detection algorithms.This method provides more security than cued click points where the user is allowed to click on a particular point called pass point for authentication which is more vulnerable to hackers