Table 1 Summarization of our proposed novel image features

1. text region entropy distribution: $H_{\mathsf{\text{h}}}^{\mathsf{\text{entropy}}}\left(\mathcal{I}\right)$, $H_{\mathsf{\text{v}}}^{\mathsf{\text{entropy}}}\left(\mathcal{I}\right)$

2. structural patterns: V_freq,h,j(j = 1, · · ·, 5), V_freq,v,j(j = 1, · · ·, 5), $H_{\mathsf{\text{h}},i}^{\mathsf{\text{structure}}}\left(\mathcal{I}\right)$, $H_{\mathsf{\text{v}},i}^{\mathsf{\text{structure}}}\left(\mathcal{I}\right)$,

3. distance distribution closest neighbours: V_dis,j(j = 0, · · ·, 4)