Features extracted from texture database after convolution with the zero and ninety degree flipped version of the original sub-mask of Discrete Cosine Transform (DCT) basis filtering masks of size 8×8 have been proposed as Rotational Invariant DCT Masks (RIDCTM) features. Based on these features query images are classified excellently by minimum distance classifier. Also proposed rotational invariant feature extraction technique has been applied to segment captured images of coal particle belonging to different category of size range. Although the proposed technique almost equals the performance of the recent rotational invariant technique based on Gabor transform in terms of classification accuracy, its efficacy lies in easier implementation and lesser computational burden like any real transform.