In the last few years biologists and pathologists are relying more and more on image analysis, and immunohistochemistry (IHC) is nowadays one of the most popular imaging techniques to analyze the presence and activity of target antigens in the tissues, with important applications in the diagnosis and assessment of tumors as well as for several research purposes. However, immunohistochemistry has been traditionally affected by lack of reproducibility due to technological variabilities as well as to the inherent subjectivity of the visual observation, thus the analysis has been limited to qualitative evaluation of the presence of the target stains within the tissues. The rapid evolution of the technique as a valid diagnostic and prognostic tool for tumor marker identification and cancer assessment has ultimately shifted the aim from qualitative to quantitative, stressing the demand for the standardization of the overall IHC assay and for the extraction of objective and repeatable measures of protein activity from the IHC images. Computer-aided image analysis has been universally acknowledged for having a fundamental role in solving the IHC standardization issue; in particular, tissue and cell segmentation techniques are precious instruments to identify the regions of interest of the target antigens in the specimens, allowing fully-automated and repeatable measurements at cellular and sub-cellular level; this is required by modern pathology and not feasible with simple visual evaluation. To this date, literature does not provide effective solutions for these challenging tasks, which gives motivation to our thesis work. This thesis addresses the problems of tissue compartmentalization and cell segmentation in IHC images, proposing fully-automated techniques to i) recognize the cancerous areas of the samples disregarding non-interesting tissues such as stroma and blood vessels; ii) detect and delineate the sub-cellular regions of the cancerous cells, i.e. nuclei, cellular membranes and cytoplasm. The proposed methods, based on color and morphological processing, were validated on large datasets of IHC cancer images from several anatomical locations and compared experimentally with state of the art segmentation approaches, such as Support Vector Machines and active contours. Our extensive experimental results demonstrate the quantitative accuracy and reproducibility of the segmentations provided by our techniques, that can be used to obtain localized measure of protein activity as well as for any other applications requiring tissue and cell exploration in pathological tissues. We conclude our thesis with final remarks about IHC quantification methods, proposing a set of requirements to obtain reliable quantifications applying computer-aided image analysis.