A first theoretical contribution consisted in extending the classical Kolmogorov- Smirnov homogeneity test to compare two samples of interval-valued observed measurements. In such a case, the test result is interval-valued, and one major diffcultyis to find the bounds of this set. We propose a very efficient computational method for approximating these bounds by using a p-box (pairs of upper and lower cumulative distributions) representation of the samples. Another contribution was the proposition of a new interval-valued fuzzy transform. Its construction is based on a possibilistic interpretation of the partition on which the fuzzy transform is built. The main advantage of this approach is that it provides specific interval valued functions whose interpretation is straightforward. This interpretation relates to a traditional sampling/reconstruction framework where little is known about the sampling and/or reconstructing kernels. Numerous properties of the proposed approach are proved that could be useful for function analysis and comparison [ART.23]. The principle of this interval-valued fuzzy transform has been adapted to the super resolution matter. Image Super-Resolution (SR) is a technique that reconstructs a high resolution (HR) image from one or several low resolution images acquired from the same scene. The most effective SR methods proposed in the literature require precise knowledge of the so called Point Spread Function (PSF) of the imager while in practice, its accurate estimation is nearly impossible. We developed a new SR method whose main feature is its ability to account for scant knowledge of the imager PSF. This ability is based on representing imprecise knowledge on the PSF by mean of an imprecise kernel. The algorithm transfers this imprecise knowledge to the output by producing an imprecise (or interval-valued) HR image. We performed some experiments illustrating the robustness of the pro-posed method with respect to. the choice of the PSF shape and parameters. These experiments also highlight its high performance compared to very competitive earlier approaches. Finally, we show that the imprecision of the HR interval-valued reconstructed image is a marker of the reconstruction error [ART.4, CACT.35].