IMPLEMENTATION OF FULLY CONVOLUTIONAL NEURAL NETWORKS (FCNN) ARCHITECTURE FOR BINARIZATION OF HISTROICAL DOCUMENT

Abstract

Binarization purpose is converting the grayscale image into binary image In the historicaldocument.
Binarization is the first step in analysing the historical document.After analyse the historical document.
The proposed system uses the neural network while the existing system was performed by pixel
classification. The Fully convolutional network (FCN) with the help of the convolutional neural network.
In this paper we propose a method to segment the document based on using dilated convolutional neural
network and pixel classification. Evaluation metric is performed by pseudo F- measure metric. The
dataset used for this paper is referred from the DIBCO.