Fibre streaks are observed in experiments with fibre suspensions in a turbulent half-channel flow. The preferential concentration methods, most commonly used to quantify preferential particle concentration, are in one dimension found to be concentration dependent. Two different new streak quantification methods are evaluated, one based on Voronoi analysis and the other based on artificial particles with an assigned fixed width. The width of the particle streaks and a measure of the intensity of the streaks, i.e. streakiness, are sought. Both methods are based on the auto-correlation of a signal, generated by summing images in the direction of the streaks. Common for both methods is a severe concentration dependency, verified in experiments keeping the flow conditions constant while the (very dilute) concentration of fibres is altered. The fixed width method is shown to be the most suitable method, being more robust and less computationally expensive. By assuming the concentration dependence to be related to random noise, an expression is derived, which is shown to make the streak width and the streakiness independent of the concentration even at as low concentrations as 0.05 particles per pixel column in an image. The streakiness is obtained by applying an artificial particle width equal to 20 % of the streak width. This artificial particle width is in this study found to be large enough to smoothen the correlation without altering the streakiness nor the streak width. It is concluded that in order to make quantitative comparisons between different experiments or simulations, the evaluation has to be performed with care and be very well documented.