Date Added: Apr 2012
Image abstraction is a method of enhancing or highlighting the visualization of main information contents of image and smoothing the other portion. It is quite important for many multimedia applications like cartooning and animation where symbolic presentation matters; hence it is an important part in the field of image processing. In this paper, the authors are proposing a neural network based approach which not only abstract the image but also avoids the enhancement of noise. This particular task is performed by empirically training the neural network for noise patterns on images by mixing the plane image with noise finally the algorithm is simulated on MATLAB and results showing that the proposed technique is fast and efficient in performing the particular task.