Multi Level Feature Priority algorithm based text extraction from heterogeneous and hybrid textual images

摘要

This paper presents a unified approach for the extraction of text from heterogeneous and hybrid textual images (both scene text and caption text in an image) and document images with variations in illumination, transformation/perspective projection, font size and radially changing/angular text. The strength of this technique lies in producing small number of features at less running time for the extraction of text from heterogeneous images in various priority levels. Proposed feature selection algorithm is evaluated with three common Machine-Learning (ML) algorithms and effectiveness is shown by comparing with three feature selection methods. The qualitative analysis proves the encouraging performance of the proposed text extraction system in comparison with the edge-, Connected-Component- (CC) and texture-based text extraction algorithm.