This paper presents an imaging technique developed to study the strain localization phenomena that occur during the tension of thin, flat steel samples. The data are processed using digital speckle image correlation to derive the two in-plane components of the displacement vectors. The authors observe that the calculation of the intercorrelation function reveals a systematic error and propose a numerical method to limit its influence. Plastic incompressibility and thin-sheet assumptions are used to derive the third displacement component and, hence, the various strain and strain rate components. Numerous checks are presented at each step in processing the data to determine the final accuracy of the strain measurements. It is estimated that this accuracy is quite sufficient to track the inception and the development of localization. Examples of possible application are presented for mild steels whose strain localization mechanisms appear to be precocious and gradual.