Old-document handwritten text extraction using machine learning

Wong, Kok Jing (2021) Old-document handwritten text extraction using machine learning. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Preview

Text (Full Text) Old-document handwritten text extraction using machine learning.pdf - Submitted Version Download (10MB) | Preview

Abstract

This project is to pre-process the captured image to filter out the background image noise without significantly remove the original text information using the machine learning. The images are mixed with four different types of noises in the process of digitization text extraction. The majority of the available algorithm can only use to remove single type of noise. For this condition, three image filtering techniques will have carried up to remove the noise that presents in the image. An ensemble model combined with the previous denoise technique is put forward for removing the noise. Meanwhile, simulation and results show that the method can effectively clear the noise. As a results， the proposed ensemble models achieved a high accuracy as 98.7 % on the Noisy Office datasets. Lastly, the results of this proposed system are better compare to the current single denoise technique and it can be implemented for segmentation process.

Actions (login required)

View Item