The author took the process of detecting signal in a noisy environment one step further by introducing the concept of optimal decision. Previously, the signal-detecting process stopped after assigning a probability p for being signal and, consequently, 1 − p for being noise, based on a statistical or optimization technique.
Using a Gaussian noise environment, the author successfully argued that a minimum for Missing Information (MI) always exists and an optimal decision can be made by minimizing this MI. However, the total MI could not be used since it “decreased without bound for a decreasing Pmin of decision.” Nor can the MI in signal be used since it did not have a minimum. But “the average MI of a noise count” could be used, since it does have a minimum. The author also claims that the procedure applies equally well to fuzzy images and other noisy environments.
This is where the reviewer has trouble agreeing with the author. I think the author was too ambitious in generalizing the procedure to fuzzy sets and other data environments. I especially object to the fuzzy image conjecture without analytical proof. Also, I have trouble accepting the author’s claim that the procedure is equally applicable to data with more signal and less data, as well as to a more complicated detection environment with parameters, such as correlation in time and space.