Vehicle License Plate Detection of Real Time Videos Using Super Resolution Techniques

The High resolution image formed using still image do miss some fine details which can increase the resolution of High resolved Image. To increase the resolution and visuality of formed high resolution image, input videos can be used. Super resolution Image recovery basically depends on two factors viz. accuracy of image registration results and restoration process where noise and blurring effects are removed. This paper presents the reconstruction of license plate using super resolution techniques from more than one low resolution view of the scene. To reconstruct a high resolution image, major challenge is to acquire accurate image registration. The displacement between the images here is calculated using affine transformations and image is reconstructed using Projection onto Convex Set (POCS) and Structure Adaptive Normalized Convolution (SANC). Experiments were implemented for super resolution techniques POCS and SANC using low resolution video. The experimental results showed the effectiveness of POCS technique over SANC technique. MSE and PSNR parameters are further used to test and compare the proposed super resolution techniques.