Fine particulate matter (PM<sub>2.5</sub>) is a major air-quality and public-health concern, while conventional monitoring networks remain limited in coverage and costly to deploy and maintain. This work presents the AQpictures project, which investigates alternative low-cost monitoring approaches based on outdoor imagery, such as webcam photographs, for PM<sub>2.5</sub> estimation. Milan, Italy, is used as a testbed for the preliminary implementation and assessment of representative image-based PM<sub>2.5</sub> estimation methods reported in the literature. The study combines a literature review with an experimental benchmark using webcam images paired with ground-sensor PM<sub>2.5</sub> measurements and ancillary meteorological data. The review confirms that a growing range of image-based PM estimation methods is emerging, although reproducibility remains limited due to the scarce availability of openly shared datasets and source code. Three modelling strategies were implemented and tested: a physics-based approach, a machine-learning regression model, and a simple deep-learning network. The tested models exhibited notable limitations overall. Machine-learning regression achieved high in-sample performance (<em>R<sup>2</sup></em> > 0.94) but generalised poorly to test data (<em>R<sup>2</sup></em> ≈ 0.21), while the other approaches showed generally low or inconsistent predictive performance. Overall, the results suggest that webcam-based PM<sub>2.5</sub> estimation can serve as a low-cost complementary approach; however, its reliability is strongly influenced by illumination conditions, dataset size, and modelling choices. These findings highlight the need for larger and more diverse training datasets, improved model robustness, and the development of standardised and openly accessible benchmarking frameworks.