4 issue 3, pp. 79 82 2015Text line extraction in unconstrained handwritten files continues to be engineering difficult problem as a result of non uniform individual scale, spatially varying text orientation, and matlab interference between text lines. In order to address those issues, which suggest engineering new cost characteristic that considers matlab interactions among text lines and matlab curvilinearity of each text line. To detect text in natural scene photos in accordance with two desktop classifiers. One is used to generate matlab candidate word regions and matlab other clear out is used to filter out matlab non text ones. In this matlab connected components in matlab images are extracted by using maximally steady external area algorithm and matlab Connected Components are then partitioned employing classifier and cluster them by using their pair wise family members.