Segmentation of patterns of curves and strokes associated with character shape geometry is a challenging task while developing Optical Character Recognition System for Indic scripts. Even though all Indic scripts are derived from a single entity, large amount of dissimilarities exist between northern and southern scripts. Complex conjunct formations add further complexity in the character shape description of the respective language. The complex properties among various scripts vary in terms of topology as well as geometry. The OCR efficiency of the these scripts depend on the effective mathematical representation with locale model as part of global features.An attempt is made in the present work in the form of extensive statistical evaluation on interrelations of isolated patterns of the script while formulating segmentation model. The canonical syllable model is proposed in the present work. The concept of meaningful unit and its applicability in the segmentation processes is proposed with a specific reference to Telugu script. Touching character segmentation is attempted while emphasising shape description.