Jsteg algorithm is vulnerable to statistical attacks because it influences the DCT histogram of an image perceptibly since it equalizes the number of paris of coefficients. F3 algorithm, instead of replacing LSB's, decrements the coefficient's absolute values. However, some shirngkage occure in this case. Beacuse 0 coefficients does not change and some 1 or -1 change to 0. Hence there is no way for receiver to know whether a 0 hide a message or not. Therefore, the messages which are embedded to 0 must be embedded again. But then, the number of even coefficients will be more than odd's, while this is not applicable for the original histogram. Does F4 (or F5) algorithm solve this problem? If so, how?
I think I found the answer (at least as theoritical). According to the paper written by Andreas Westfield, F4 and F5 solve this problem by mapping even negative coefficients to a steganographic, odd negative coefficients to a steganographic zero. Through this map, not only embedding 0's cause to shrinkage but also embedding 1's. Thus the shape of histogram will be preserved.