The process of the characteristic feature extraction is one of the adopted techniques for the purpose of pattern recognition in the images. The process of the characteristic feature extraction numerically is one of the adopted techniques for the purpose of pattern recognition in the digital images, and the process of the characteristic feature extraction depending on the Co-occurrence matrices is one of the most important techniques for the purpose of pattern recognition in the textured images. This work aim to study the characteristic features for the textured images, eight characteristic features are selected to investigate the aim of this work. These selected features are: Maximum probability, entropy, homogeneity, cluster shade, cluster prominence, contrast, angular second moment, and the inverse difference moment. In this research, the characteristic features depending on the Co-occurrence matrix are extracted in two ways. In the first one, the characteristic features are extracted depending on average Co-occurrence matrices which be extracted for four angles (0o, 45o, 90o, and 135o). While in the second method, the characteristic features are extracted depending on the Co-occurrence matrix for each angle of the following angles (0o, 45o, 90o, and 135o). In this method, four values for each of the selected characteristic features are extracted. Then the average values for each of the characteristic features are extracted depending on the extracted four values. To study the effect of block size on the calculation of the statistical characteristic features, the statistical features are calculated for the whole image and for each block in the image after dividing the image into blocks with block size (32x32) and for each block in the image after dividing the image into blocks with block size (64x64). In addition, to study the effect of quantization level on the calculation of the statistical characteristic features three values (8, 16 and 32) of quantization level are adopted in this research. All the calculations are applied on the three textured images with 256 gray levels selected from Brodatz album. The results show the calculation for most the selected features not change except the feature of the entropy where the difference in the extracted value of the entropy in the two ways is perceptible. This property can be utilized to increase the discrimination power in the classification process.