I am trying to implement an IEEE paper based on edge adaptive steganography. For that I have to convert a $$m \times n$$ resolution image to a row vector by raster scanning. Then I have to take two consecutive pixels and of the form $$(x_{i}, x_{i+1}) \textrm{ } \forall \textrm{ } i \in \textrm{ } 1, 3, 5, ..., mn -1$$ assuming that $$n$$ is even. Now I know the importance of $$n$$ being even. But if $$n$$ does become some odd number then what should be my approach? Should I take the last pixel which is left alone with my own assumed pixel value, say, 0? Will that work, or something else is needed?