We have extracted and stored character features SCF from all character images. Extracted a feature from test image of Urdu text line, afterwards, taking a smaller sample calculated maximum likelihood to match it against each SCF in inference algorithm and continue until whole sample is consumed, revealing all the characters in it. The algorithm may reduce complexity using smaller dictionary of Urdu characters instead of using a very larger set of ligatures. This work can be used in problems like OCR system, number plate segmentation and postal address processing, form processing etc.