Abstract The objective of this doctoral thesis has been the development of complete methodologies for the experimental characterization of the processes of fuel injection, vaporization and mixture formation, autoignition and combustion in real multi-orifice nozzles for Diesel engines. Different visualization techniques have been implemented to achieve this objective, and algorithms for automatic image processing have been developed. The study of liquid spray visualization, in experimental facilities with one or more wide optical accesses, has focused on the analysis of the effect of the opto-electronic configuration upon the image quality, and new processing algorithms have been designed and evaluated. As an innovative contribution, the LRT segmentation algorithm has been generalized to non-Gaussian distributions. Schlieren and Shadowgraph techniques have been implemented for the study of vaporizing sprays making use of two basic variants, according to the availability of optical accesses: transmission method and double-pass method. A complete theoretical-experimental study has been carried out, with the aim of defining the variables that affect the sensitivity of the method, and to identify error sources in the image analysis. Proposed algorithms are based upon the liquid spray segmentation ones, with small modifications for the calculation of the threshold digital level. The study of autoignition and combustion has been carried out by means of the recording of natural chemiluminescence emission of CH and OH radicals, as well as of the soot thermal radiation. The sensitivity of the visualization method to the acquisition equipment configuration has been analysed in detail, and an image analysis methodology based upon the statistical study of spatial intensity distribution has been presented.