Sport biomechanics studies the movements realized by athletes to be used in kinematic analysis. These analysis involve video sequences, and provide a tool for coaches to help in performance improvement and injury prevention. In particular, motion analysis in swimming is useful for improving the athletic gestures, reduce friction with water and reduce times. Traditionally, video sequences are combined with gyroscopes and accelerometers in motion analysis. However, these methods have limitations associated with portability, since they require electronic interfaces and use invasive elements for swimmers. In addition, most state of the art techniques for motion analysis are not focused on watersports, or rely on swimmers performing the movements out of the water. Additionally given that traditional algorithms have not treated the waves and turbulence produced by swimmers, and the predominance of blue and green hues make the problem more challenging. This paper presents an algorithm that processes underwater video sequences for swimmers detection and tracking using light absorbance in conjunction with compressive sensing concepts. The proposed algorithm was tested using two video sequences with different characteristics. Results show that 89.5% of the tested cases were successful.