Pixel-level Reconstruction and Classification for Noisy Handwritten Bangla Characters

TL;DR

This paper presents a novel pixel-level denoising and classification method for noisy handwritten Bangla characters using deep belief networks and probabilistic quadtrees, improving recognition accuracy.

Contribution

It introduces a pixel-level classifier with deep belief networks to effectively remove noise and enhance classification of handwritten Bangla characters.

Findings

Effective noise removal from handwritten characters

Improved classification accuracy on Bangla datasets

Demonstrated robustness to noisy inputs

Abstract

Classification techniques for images of handwritten characters are susceptible to noise. Quadtrees can be an efficient representation for learning from sparse features. In this paper, we improve the effectiveness of probabilistic quadtrees by using a pixel level classifier to extract the character pixels and remove noise from handwritten character images. The pixel level denoiser (a deep belief network) uses the map responses obtained from a pretrained CNN as features for reconstructing the characters eliminating noise. We experimentally demonstrate the effectiveness of our approach by reconstructing and classifying a noisy version of handwritten Bangla Numeral and Basic Character datasets.