Rendering of Impaired Visual Effects on Genesis of Streak - Recognizing Car License Plate

摘要

Rain Streak (RS) makes the recognition of License Plate (LP) more difficult as it partially hides the alphanumeric characters in the LP by creating patches on the LP. The high velocity rain drops over an image or video creates an undesirable interference. In real time, this streak is a form of occlusion which results in imperception of detecting the required candidate region like contours, boundary region etc. This results in impaired visual effects. Detection and removal of RS is a challenging task. It is observed that rain streak removal on License Plate (LP) results in loss of information. This paper provides an overview of rain streak removal schemes and implementation results of rain streak detection using guided filter and removal using Dictionaty learning. From the results, it is found that the rain streak removal creates impairment in License Plate alphanumeric character recognition. To rebuild the alphanumeric characters, context based character recognition using CNN can be attempted.