This study proposes the use of features combination and a non-linear kernel to improve the classification rate of texture recognition. The feature vector concatenates three different sets of feature: the first set is extracted using grey-level cooccurrence matrix, the second set is collected from three different radii of local binary patterns, and the third set is generated using Gabor wavelet features. Gabor features are the mean, the standard deviation, and the skew of each scaling and orientation parameter. The aim of the new kernel is to incorporate the power of the kernel methods with the optimal balance derived from the features. To verify the effectiveness of the proposed method, numerous techniques are tested using the three data sets, which consist of various orientations, configurations and lighting conditions.