Comparative Analysis of Multi-scale Wavelet Decomposition and k-Means Clustering Based Text Extraction

Abstract

Text present in images provides important information for automatic annotation, indexing and retrieval. Therefore, its extraction is a well-known research area in computer vision. However, variations of text due to differences in orientation, alignment, font, size, contrast and complex background makes the problem of text extraction extremely challenging. In this paper, we propose an efficient method to extract text regions even under complex background using DWT and k-means clustering along with voting decision process. As textures of text have abrupt variation and irregular texture property in the wavelet transform domain, so wavelet transform seems to be the best choice for achieving the objective of image segmentation. A small size overlapping sliding window is used to scan high frequency component sub-bands from which texture features are extracted. On the basis of these features, k-means clustering is employed to classify the image into text and background clusters. Finally, voting decision process and area-based filtering are used to locate text regions accurately. We examined and evaluated the performance by varying wavelet functions and decomposition levels. The proposed method is evaluated on four standard datasets (ICDAR 2013, KAIST, MSRA-TD500, SVT) and own created dataset. Further, performance analysis reveals that this method is robust and efficient for extracting text regions under various conditions.