Research on multi-channel spectral prediction model for printed matter based on HMSSA-BP neural network

Accurate color reproduction requires completely faithful to the original and improve prints quality. To achieve this, the pre-press color prediction is particularly critical. However, traditional Chromaticity-based prediction fades in metamerism problem inevitably. Moreover, there is no single quality index that significantly outperforms others or provides the best performance in all situations. To overcome it, this paper proposes a multi-channel spectral prediction model for printed matter and the adaptive evaluation method based on multi-index fusion. First, BPNN prediction model utilizes multi-light sources multi-spectral imaging technology, mapping the dot area ratio of C, M, Y, K to the multi-channel spectral images. Second, Hybrid Multi-strategy Sparrow Search Algorithm (HMSSA) is constructed to optimize BPNN, which combines Tent mapping, step size phased control, and chaotic cosine transform factor. Third, multi-channel spectral images synthesis method which introduces adjusting factor Q, obtains the better predicted color image. Then, adaptive evaluation model of multi-index fusion is built to evaluate the predicted image quality, including SSIM, Spearman’s coefficient, Bhattacharyya distance, and PSNR. Several experiments are performed to verify the significance of the proposed method under different scenarios. Compared with the existing methods, the proposed multi-channel spectral prediction model exhibits the superiority in improving the accuracy of predicting the actual printing image.