This paper presents DWT based text extraction techniques from various types of images. Text data present in images and videos contain useful information for data retrieval, annotation, indexing and content analysis. This paper discusses various methods such as edge detection filter, discrete wavelet transform (using Haar, Daubechies and Splines), dual-tree discrete wavelet transform (DT DWT), segmented-matrix algorithm for Haar DWT on Central Processing Unit (CPU) and Graphical Processing Unit (GPU) etc. Text extractions in images are used in variety of applications such as visually challenged person, vehicle license plate detection and bank cheque processing etc. Variations of text in images occur due to font, font size, orientation, languages, styles and shapes. This paper provides performance comparison of several existing methods suggested by researchers in DWT based text extraction on the basis of recall rate, precision rate, processing time, detection rate and false alarm rate etc.
Artificial Neural Network (ANN), Central Processing Unit (CPU), Discrete Cosine Transform (DCT), Discrete wavelet Transform (DWT), Dual-Tree Discrete Wavelet Transform (DT DWT), Graphical Processing Unit (GPU), Kurtosis, Wavelet Modulus Maximum (WMM)
