Frequency domain analysis is the background of representation of the feature vector. Different textural and statistical values are also computed which enriches the feature vector. Entropy, which is a statistical measurement of the randomness of an image, is computed on grey-scale, binary, and twelve sub-band images (for \(\phi_{1}\), \(\phi_{2}\) and \(\phi_{3}\)). Similarly, statistical measures like mean, standard deviation are also computed on the same set of images to enrich the proposed feature vector \cite{obaidullah2015numeral}.

The generation of the features vector is composed by 55 features described in the following way  \cite{obaidullah2015numeral}:

Texture analysis on gray-level and binary image is given the first feature (\(F_{1}\)). By applying statistical features like entropy (\(F_{2}\)), mean (\(F_{3}\)) and standard deviation (\(F_{4}\)). This generate four features \(F1\) to \(F4\). By wavelet decomposition approximation coefficient \(A_{1}\) are given en the fist stage \(\phi_{1}\), horizontal coefficient \(H_{1}\), diagonal coefficient \(D_{1}\), vertical coefficient \(V_{1}\) are also computed in this stage. The entropy, mean and standard deviation are calculated on each of the decomposed sub-images generating twelve features \(F_{5}\) to \(F_{16}\). Wavelet entropy using five entropy types namely \(shannon\), \(logenergy\), \(threshold\), \(sure\) and \(norm\), which are computed on approximation coefficient \(A_{1}\) sub-image. This process will generate five features \(F_{17}\) to \(F_{21}\). Finally, this process is repeated for linear an quadratic coefficients for the second and thirth stage of the decomposition wavelet (\(\phi_{2}\) and \(\phi_{3}\)), generating the 17 of the rest features for each one. That means: \(F_{22}\) to \(F_{38}\) for \(\phi_{2}\) and \(F_{39}\) to \(F_{55}\) for \(\phi_{3}\).

The proposed method by  \cite{obaidullah2015numeral} generates a feature vector of 55 where \(F_{1}\) is done by \(m/2\) \(*\) \(n/2\) attributes given by the size of the image at the first stage \(\phi_{1}\).