Steep mountain rivers have hydraulic and morphodynamic characteristics that hinder velocity measurements. The high spatial variability of hydraulic parameters, such as water depth (WD), river width and flow velocity, makes the choice of a representative cross-section to measure the velocity in detail challenging. Additionally, sediment transport and rapidly changing bed morphology exclude the utilization of standard and often intrusive velocity measurement techniques. The limited technical choices are further reduced in the presence of macro-roughness elements, such as large, relatively immobile boulders. Tracer tracking techniques are among the few reliable methods that can be used under these conditions to evaluate the mean flow velocity. However, most tracer tracking techniques calculate bulk flow velocities between two or more fixed cross-sections. In the presence of intense sediment transport resulting in an important temporal variability of the bed morphology, dead water zones may appear in the few selected measurement sections. Thus a technique based on the analysis of an entire channel reach is needed in this study. A dye tracer measurement technique in which a single camcorder visualizes a long flume reach is described and developed. This allows us to overcome the problem of the presence of dead water zones. To validate this video analysis technique, velocity measurements were carried out on a laboratory flume simulating a torrent, with a relatively gentle slope of 1.97% and without sediment transport, using several commonly used velocity measurement instruments. In the absence of boulders, salt injections, WD and ultrasonic velocity profiler measurements were carried out, along with dye injection technique. When boulders were present, dye tracer technique was validated only by comparison with salt tracer. Several video analysis techniques used to infer velocities were developed and compared, showing that dye tracking is a valid technique for bulk velocity measurements. RGB Euclidean distance was identified as being the best measure of the average flow velocity.