Laser-beam induced current imaging is a powerful technique for assessing the spatial performance of photovoltaic devices. However, it has not enjoyed widespread industrial application since high resolution measurements of large samples can last several hours, while weak signals from micrometre sized laser spots require lock-in amplification. The technique usually requires a modulated laser beam scanning the device’s surface, measuring the corresponding photocurrent before moving to the next point and repeating. An alternative approach presented recently is the use of digital micromirror devices combined with compressed sensing theory. There are significant benefits when using such methods, such as signal amplification, undersampling and simplified measurement systems. These benefits lead to a more sensitive and agile inspection tool. Nevertheless, high computational requirements have limited the experimental application of this method to low resolution outputs. In this work we present the mathematical background and the experimental approach towards megapixel resolution, ultrafast compressed sensing current mapping, overcoming previous computational barriers. In addition, a high power Digital Light Processing projection system is introduced for the experimental application. Multilevel sampling using structured patterns is presented and the wavelet transform is adopted as an efficient approach for reliable reconstruction results for compressed sensing current mapping. Solutions to computational issues, sampling optimisation and measurement strategies are presented. The projection-based measurement acquisition provides great flexibility regarding the sizes of samples that can be measured, while it accelerates measurement speed by an order of magnitude. The proposed methodology is illustrated through high resolution experimental results and close to megapixel current maps of up to wafer-sized cells are presented.