Mixed Gaussian-Impulse noise robust face hallucination via noise suppressed low-and-high resolution space-based neighbor representation

An intelligent surveillance system poses a lot of challenges in the processing of captured noisy low-resolution (LR) images. To defeat such challenges, face super-resolution (SR) also called face hallucination techniques are getting prominence in recent years. Although, the present SR models are not good enough to handle the complicated noise e.g., mixed Gaussian-Impulse (MGI) noise, often present in the captured LR images. Therefore, a new MGI noise-robust face hallucination algorithm using noise suppressed low-and-high resolution space-based neighbor representation (NSLHNR) is proposed in this paper. The proposed algorithm first suppresses the effect of outliers from the SR process by overlooking them from the reconstruction weight calculation process. It assists in controlling the square reconstruction error. Further, it also accomplishes the HR space-based neighbor representation to counterbalance the losses caused due to high-density MGI noise in a relationship of input and training LR images. These additions make the proposed algorithm capable to preserve sharp edges, texture, and the individual characteristics of the input face in the output. The performance measured through the experiments performed on the benchmark datasets and surveillance images shows the better reconstruction capability of the proposed algorithm over the compared state-of-the-art models.

Automated summary

An intelligent surveillance system faces challenges in processing noisy low-resolution images, leading to the development of face super-resolution techniques. The proposed algorithm effectively suppresses noise and preserves key features of input faces, showing better reconstruction capability compared to state-of-the-art models.