抄録
An Optimal black separation signal prediction method in the mean square error (MSE) sense is presented for color printing image coding. In this method, a prediction error signal (eK) is defined as the difference between an actual black separation k signal and a predicted k signal with luminance. In this prediction, the k signal predictor is statistically designed under a criterion of minimization of the mean square prediction error, so that the predicted value is equal to the conditional expected value of k for each luminace level. By the cmy-YIQ conversion and the use of an eK signal instead of a k signal, the spectral redundancy among the c, m, y and k signals is removed, and more efficient coding is expected. In the coding scheme, these Y, I, Q and eK signals are spatially coded separately. The simulation results show that YIQeKDPCM coding applying this prediction method is superior to conventional YIQkDPCM coding in SNR for the k signal.
