22 Mar 2024
10:30 Master's Defense fully remotely
Theme
Unsupervised Deep Machine Learning Method for Haze Removal without Paired Images
Student
Percy Maldonado Quispe
Advisor / Teacher
Helio Pedrini
Brief summary
In this work, we address a fundamental and still relatively little explored aspect in the field of artificial neural networks for unsupervised image dehazing. By conceiving a hazy image by superimposing several simpler layers, such as a haze-free image layer, a transmission map layer and an atmospheric light layer, inspired by the atmospheric scattering model, we propose an approach based on the concept of layer untangling. Our method, called XYZ, represents a substantial improvement in image quality metrics such as SSIM and PSNR, as well as BRISQUE, PIQE and NIQE. This advancement is achieved through the strategic combination of the XHOT, YOLY and ZID methods, capitalizing on the individual strengths of each. A distinctive and valuable aspect of the XYZ approach is its unsupervised nature, which implies that it does not rely on datasets containing pairs of sharp and blurred images for training. This contrasts with the traditional deep training paradigm, marking an innovation in the field of haze removal. Furthermore, we highlight two fundamental benefits of the proposed XYZ approach. Firstly, because it is unsupervised, it frees you from the need to use exhaustive datasets that include sharp and blurred images as a fundamental reference. Secondly, we approach the issue of haze from a multifaceted perspective, recognizing and unraveling the complexities inherent in this atmospheric phenomenon. This layered approach allows for a more accurate and detailed representation of the scene, thus improving the quality of fog-free images.
Examination Board
Headlines:
| Hélio Pedrini | IC / UNICAMP |
| Helena de Almeida Maia | IC / UNICAMP |
| César Armando Beltrán Castañón | PUCP |
Substitutes:
| Andre Santanche | IC / UNICAMP |
| Ronaldo Cristiano Prati | CMCC / UFABC |