Edge Enhancement Using Co-Occurrence Features of LBP Coded Low Contrast MR Images

Restoration of image features from local binary pattern (LBP) coded MR images form an important application in Telemedicine. LBP is invariant to monotonic gray scale change and it can be used as a rotational invariant texture descriptor. In this paper, we use co-occurrence features of local binary patterns to edge enhance low contrast MR images. It is thereby shown that the co-occurrence features are able to retrieve the anatomical information present in the original MR image, which otherwise can be obtained using a standard edge enhancing kernel. Out of 14 textural features extracted from co-occurrence matrix, the measure sum-average performs better in preserving anatomical information.