A comparative Analysis of Feature Extraction Algorithms and Deep Learning Techniques for Detection from Natural Images
Deepak Kumar, Ramandeep Singh · 2019
Text detection from natural scene images have major role in image content analysis. It is considered as complex issue because of the variability in text size, fonts, complex background, line orientation in image with non-uniform illumination. For overcoming the issue, efficient text image recognition features are being utilized. This research has proposed the comparative analysis of various text detection technique based on pixel dependable feature extraction and deep learning dependable text region classification. The feature extraction technique includes extraction of feature sets from the text region which divides input natural image into several regions. The text regions are identified using morphological operations like binarization, thinning, segmentation etc. based on the color, consistency and numerical features. In this paper, the effectiveness of feature extraction algorithms and deep learning techniques are compared on the basis of various performance parameters to detect the text from natural images.