Dominating set based arbitrary oriented bilingual scene text localization

<p>Localizing and recognizing arbitrarily oriented text in natural scene images is the biggest challenge. It because texts are often erratic shapes. This paper presents a simple effective graph representational algorithm for detecting arbitrary-oriented location to smoothen recognition process of its high impact simplicity representation. An can be horizontal, vertical, perspective, curved (diagonal/off-diagonal), or even combination. As pre-processing step, image enhancement performed frequency domain improve representation that invariant intensity. necessary draw bounding boxes each candidate character extract regions. step carried out by utilizing advantage region-based approach called maximally stable extremal A typical problem with localization non-text objects may occur within localized Our method first literature searches dominating sets solve this problem. set outperforms several traditional methods, including deep learning methods used arbitrary localization, on challenging datasets like 13<sup>th</sup> international conference document analysis (ICDAR 2015), multi-script robust reading competition (MRRC), CurvedText 80 (CUTE80), (ArT).</p>