It is evident to see that points of gray level images cannot be characterized in an accurate way only by using derivatives of the first order, since it is reduced to two invariants: the gray value of the pixel and the magnitude of its gradient. Second and may be third order invariants, as Schmid shows it in [14], must be used to obtain interesting results. But this characterization is quite expensive to implement and may be not very sure, according to the fact that it is difficult to compute the second and third derivatives in a stable way. The color information, by multiplying the number of invariants by three and by adding two invariants specific to color, allows us to characterize points with accuracy and simplicity, only by using the derivatives till order one.