فاطمه فلاحي

By focusing in photography, we can adjust the focal length of the camera lens based on the distance from the object to the camera to get a sharp image. Achieving proper focus in images has become very important by using focus distortion correction methods. In this thesis, we have discussed focus and the possibility of using deep learning method to remove focus distortion from images. To this end, a dataset including images without focus distortion and images with focus distortion has been produced for training the network. The results show that after the learning process, the designed neural network will be able to remove focus distortion in images with more than 82% accuracy. The obtained values for the image quality eva‎luation parameters and the accuracy of the network in removing the distortion show the efficiency of the network in detecting and correcting the focus distortion, and surpasses the conventional methods with limitations in the automatic extraction of features and access to large data sets.

فوكوس , پراش فرنل , شبكه عصبي كانولوشن , شبكه عصبي رمزگذارخودكار , يادگيري عميق , هوش مصنوعي , تصاوير فازي

focus , Fresnel Diffraction , convolutional neural network , Autoencoder Neural Network , Deep Learning , artificial intelligence , phase images

Fateme Fallahi

Dr. Mohammad Reza Jafarfard