This paper proposes a new descriptor for texture classification that is robust to image blurring. The descriptor utilizes phase information computed locally in a window for every image position. The phases of the four low-frequency coefficients are decorrelated and uniformly quantized in an eight-dimensional space. A histogram of the resulting code words is created and used as a feature in texture classification. The method is highly insensitive to blur, as the low-frequency phase components are invariant to centrally symmetric blur. The method is also invariant to uniform illumination changes. Experiments show that the new method achieves higher classification accuracy for blurred texture images compared to well-known methods like LBP and Gabor filter banks. Interestingly, it also performs slightly better for non-blurred textures. The method, called local phase quantization (LPQ), is based on the blur invariance property of the Fourier phase spectrum. It uses the local phase information extracted using the 2-D DFT or a short-term Fourier transform (STFT) computed over a rectangular window at each pixel position. Only four complex coefficients are considered, corresponding to 2-D frequencies. These coefficients are decorrelated and quantized, resulting in an 8-bit code. A histogram of these codes is used as a feature vector for classification. The LPQ method is highly tolerant of blur, as the phases of the low-frequency components are ideally invariant to centrally symmetric blur. Although the invariance is disturbed by the finite-sized image windows, the method is still very tolerant of blur. The method is also invariant to uniform illumination changes. Experiments show that the LPQ method achieves higher classification accuracy for blurred textures compared to other methods. It also performs slightly better for sharp textures. The results demonstrate that the LPQ method is effective for texture classification in the presence of blur.This paper proposes a new descriptor for texture classification that is robust to image blurring. The descriptor utilizes phase information computed locally in a window for every image position. The phases of the four low-frequency coefficients are decorrelated and uniformly quantized in an eight-dimensional space. A histogram of the resulting code words is created and used as a feature in texture classification. The method is highly insensitive to blur, as the low-frequency phase components are invariant to centrally symmetric blur. The method is also invariant to uniform illumination changes. Experiments show that the new method achieves higher classification accuracy for blurred texture images compared to well-known methods like LBP and Gabor filter banks. Interestingly, it also performs slightly better for non-blurred textures. The method, called local phase quantization (LPQ), is based on the blur invariance property of the Fourier phase spectrum. It uses the local phase information extracted using the 2-D DFT or a short-term Fourier transform (STFT) computed over a rectangular window at each pixel position. Only four complex coefficients are considered, corresponding to 2-D frequencies. These coefficients are decorrelated and quantized, resulting in an 8-bit code. A histogram of these codes is used as a feature vector for classification. The LPQ method is highly tolerant of blur, as the phases of the low-frequency components are ideally invariant to centrally symmetric blur. Although the invariance is disturbed by the finite-sized image windows, the method is still very tolerant of blur. The method is also invariant to uniform illumination changes. Experiments show that the LPQ method achieves higher classification accuracy for blurred textures compared to other methods. It also performs slightly better for sharp textures. The results demonstrate that the LPQ method is effective for texture classification in the presence of blur.