This study explores the image analysis practices of undergraduate translation trainees in a Greek. university. The key research questions focused on the trainees' perception of the role of visual literacy in translation, the nature of their verbal and visual associations and activities effective for developing this role. Visual literacy, semiotics, and translation theories provided the conceptual framework for this multimodal semiotics approach to translation training, which guided my examination of the way the trainees described semiotic resources, how they used verbal and visual elements of Cl. translation task, and the extent to which they focused on the visual. Expanding on the concept of mediation, I developed a 'verbo-visual' mediation model that included visual semiotic elements to provide a detailed description of the ways in which the trainees intervened in the translation process. Through an action research study I gauged the extent to which the trainees had developed visual literacy skills. I planned a series of tasks centred on the translation of a multimodal text from English into Greek. I used direct and video observation of the students as they performed the tasks and afterwards evaluated their translations. The main findings of my research reveal that the trainees found it difficult initially to consider all the semiotic elements of the data-text whilst simultru1eously translating its verbal elements. However. by providing examples of potential verbal-visual interactions and reflecting on and modifying the tasks, I found that the trainees began to appreciate non-verbal elements as potential translational factors. I was thus able to demonstrate a raised level of awareness of the visual elements of multimodal texts to some extent for all trainees in this study. The research suggests that customized image analysis, photo-elicitation and visualization techniques may play a part in the improvement of translation training in a multimodal semiotics translation environment.