Symmetrical-symbol pattern based structured light for accurate decoding

In this paper, we propose a set of four/eight symmetrical symbols for structured light pattern generation. Four preliminary steps of decoding the pattern are proposed namely (1) pre-processing, (2) classifying symbols, (3) finding neighbors and (4) matching and correction. Two additional techniques to decode a symbol located at the border of the pattern and to correct a wrong codeword are also introduced and applied. Experiments are done for three surface types including plane, curve and folded paper. Decoding border symbols helps to increase the number of decoded symbols by 17.97%, 16.68% and 15.25% over 50 samples of plane, curve and folded paper, respectively. After applying both of the two techniques, the ratios of decoded symbols to the total of detected symbols are 100%, 100% and 97.2%, which are 0.3%, 4.9% and 0.23% higher than those of the existing methods, respectively.