Handwritten Numeral Recognition using Local Intensity Order Pattern of Popular South Indian Scripts

Multi-lingual multi-script often poses various challenges in handwritten numeral classification. In this paper, we present a method for classification of off-line handwritten numerals of three popular Indian scripts. Here we consider Kannada, Telugu and Tamil scripts for our experiment. The features used in the classifier are obtained from local intensity order patter descriptor. For feature computation, Local Intensity Order Pattern (LIOP) is used. Basically LIOP encodes the local ordinal information of each pixel and the overall ordinal information is used to divide the local patch into sub-regions which are used for accumulating the LIOPs respectively. Therefore, both local and overall intensity ordinal information of the local patch are captured by the LIOP descriptor so as to make it a highly discriminative descriptor. LIOP descriptor features are not only invariant to monotonic intensity changes, image rotation but to many other geometric and photometric transformations such as viewpoint change, image blur issues are effectively addressed in our experiments [7]. A five-fold cross validation technique with numeral network classifier has been used for result computation and we obtained 98.34%, 98.40%, and 97.51% accuracy for Kannada, Telugu, and Tamil scripts respectively.