Coding of correlated sources with prescribed distortion by separated encoders.
Let (X(n),Y(n)) = (X(1),Y(1)),...,(X(n),Y(n)) be n independent, identically distributed pairs of chance variables. Each of two separated encoders encodes a different sequence of (X(n),Y(n)) to within given respective distortions z(1),z(2). Call a region in the plane of the respective rates (R(1),R(2)) M-convex if it is convex, closed, and, to every point (r(1),r(2)), contains all points such that each of their coordinates is larger than or equal to the corresponding coordinate of (r(1),r(2)). This paper determines the (M-convex) region of achievable rates by giving its boundary. For z(1) = z(2) = 0, the region specializes to that of Slepian and Wolf. For z(2) = infinity and R(2) = H(Y(i)), the minimum R(1) is that of Wyner and Ziv; for z(2) = infinity and general given R(2), the minimum R(1) is that of my generalization of the theorem of Wyner and Ziv.