A modification of the two-flux Kubelka-Munk (K-M) model was proposed to describe the energy conservation of scattered light in colored mixed material with a defined scattered photometric, which is applied for the relative quantity distribution of each colored monochrome component in mixed material. A series of systematical experiments demonstrated a higher consistency with the reference quantity distribution than the common Lambert-Beer (L-B) law. Its application in the fibrogram of each component for measuring the cotton fiber’s length was demonstrated to be good, extending its applicability to white and dark colored blended fibers, the length of which is harder to measure using L-B law.

Fiber length is one of the essential and tricky-to-get characteristics of colored cotton fibers, which affects the spinnability and quality of blending yarns. Based on the optical theory and structure traits of fiber assemblies, an optical method named Random-Beard Image Method (RBIM) was proposed for measuring linear density distribution, fibrogram, of dyed and natural colored cotton fiber beards. This method involves a new two-level prediction model of reflectance and an optical algorithm for beard bulkiness. Fibrograms of 22 different cotton samples in total were tested to verify such algorithms and compared with those by the international standard method, Advanced Fiber Information System (AFIS). The average error between RBIM and AFIS was turned to be lower by 3.22%, and an excellent agreement of fibrograms between the two methods was also confirmed by Bland-Altman analysis. Meanwhile, an exceptionally high linear correlation on the average weight length of these two methods was achieved with insignificant deviation under a significance level of a = 0.01 with the T-test.

Insufficient studies of reflectance and fluffiness effects on optical density caused large derivation and limited the application of fibrogram methods in colored cotton fiber length tests for decades. A coupled optical method was proposed here to solve this issue, named reflectivity and fluffiness coupled optical algorithm of the random-beard image method. In this algorithm, a defined theoretical reflectance was declared with an iteration method for optical linear density on the optical analysis of the derived Kubelka–Munk law, which satisfied its assumption of excluding reflected light at the interface with air. Meanwhile, an apparent polynomial relationship was revealed between the actual density and optical densities of fiber assemblies at different fluffiness. Its accuracy on weighted mean fiber length was demonstrated, referred to as advanced fiber information system, using 10 kinds of dyed cotton slivers and 12 kinds of naturally colored lint fibers whose mean and maximum absolute errors are 0.59 mm and 1.6 mm, respectively. Furthermore, their excellent agreement on fibrograms and the weighted mean length was confirmed by Bland–Altman analysis, correlation analysis and the T-test. Finally, an exceptionally high linear correlation (r2 = 0.93) was achieved on their weighted mean length and had no statistical difference with insignificant deviation under a significance level of a = 0.01. This new algorithm, featured fast and low cost, could provide accurate mean fiber length of colored or dark cotton, exhibiting valuable guidance in the potential application of the fiber trade and the quality control of colored cotton fibers or primaries in blending assemblies and yarn.