Robust local ternary patterns for texture categorization
Xian‐Hua Han, Gang Xu, Yen‐Wei Chen · 2013
This paper proposes a new image representation for texture categorization, which is based on extension of local binary patterns (LBP). As we know LBP can achieve effective description ability with appearance invariance and adaptability of patch matching based methods. However, LBP only thresholds the differential values between neighborhood pixels and the focused one to 0 or 1, which is very sensitive to noise existing in the processed image. This study extends LBP to local ternary patterns (LTP), which considers the differential values between neighborhood pixels and the focused one as negative or positive stimulus if the absolute differential value is large; otherwise no stimulus (set as 0). With the ternary values of all neighbored pixels, we can achieve a pattern index for each local patch, and then extract the pattern histogram for image representation. Experiments on two texture datasets: Brodats32 and KTH TIPS2-a validate that the robust LTP can achieve much better performances than the conventional LBP and the state-of-the-art methods.