Face image processing by TV filter and super-resolution

The demand for human face enhancement in pictures is increasing. This paper describes an effort to utilize state-of-the-art signal processing technologies for the enhancement of the human face in pictures. First, several nonlinear filters are examined, and it is demonstrated that the total variation regularization filter (TV filter) shows the remarkably best effect for skin smoothing including the removal of wrinkles, stains, moles, and freckles. The reason is analyzed in detail. Then, super-resolution technology is utilized to enhance the image quality for specific parts of the face, such as the eye line, pupil, eyelashes, and hair. Facial part extraction technology is also utilized for the enhancement of selected face parts. Interestingly, we found that the super-resolution technology not only improves the clarity of the image but also increases the brilliancy in the pupil and hair. The super-resolution technology used in this paper is based on the non-linear filtering method developed for 4K high-definition television.