Abstract
Super Resolution (SR) algorithm produces a high resolution (HR) image from single or multiple low resolution (LR) images. This algorithm is used to overcome the limitation of imaging CMOS sensors. It is difficult to obtain a HR image by reducing the size of the sensor after a certain limit. SR technique is used in many visual applications like biological imaging, military applications and forensic investigations. It is basically an inexpensive process to enhance the resolution of an image and to extract the high-frequency information. Two different adaptive schemes are proposed here. First one focuses on minimizing the error between the actual image and the estimated image. Resolution enhancement is done here by simultaneously modeling a blurring filter to capture the degradation process as well as modeling an innovation filter to remove the blurring effects, sensor noise using adaptive Least Square technique. The second scheme incorporates the advantages of both visual quality improvement as well as the increase in PSNR by jointly using wavelet transforms and adaptive normalized Least Mean Square (NLMS) technique. Results and performances of these novel techniques are compared with other available Super Resolution methods in terms of the visual quality index like PSNR, SSIM. Numerical results indicate that these computationally efficient single image super resolution techniques are very effective in real life imaging applications as a significant improvement of visual quality is observed in the super resolved image.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have