Abstract:
Image represents the external form of an object in the form of art. In digital technology
depended world, image processing is a vast research area. Image processing is one of the most essential part of Computer Vision. Face Recognition and detection is another
beneficial part of this area. Lately, reference-based face restoration techniques rose up
highly and on great talk among world-wide researchers because of its immense potential
in resolving high density details over low resolution images. Nonetheless, most of these
approaches have limitations in requirements which is they need high standard trained
image of similar identity. Besides, many analysts have presented methods and algorithms
to solve misfocus, motion blur. But most of them performed over the whole picture pixel
and so can’t detect the main goal of the image mostly. To address the issue of restoring
face from blur surrounding, in this study we applied the accurate reference-based deep face
dictionary (DFDNet) algorithm. In this algorithm, four steps are performed to reach the
expected output of restored image along with face detection. Here, in this method,
upscaling of picture is done to repair each and every pixel inside an image in details. This
method works by feature matching of input image to pre-coached high quality images
reference dataset and at the very end, a good output is established. This algorithm works
both over synthetic and real time pictures but don’t need any personal information from
the image. Side by side, we compared other techniques and algorithms with our applied
algorithm where it is effectively found that this algorithm can reach promising outcomes
on debased low resolution images.