Optimization of Digital Media Art Creation Process Based on Dual Transformer Residual Network (DTRN) and Lotus Effect Optimization Algorithm (LEA)

Abstract

In response to the problems of lack of intelligence, low recognition, loss of brightness, and unclear design objects in the comprehensive material painting of installation art, when materials enter the field of artistic creation, they can assist in artistic creation, but constantly updating materials is also a challenge for artists. This manuscript proposes an image processing technology in the digital media art creation based on dual transformer residual network (DTRN) and lotus effect optimization algorithm (LEA) (DMAC-DTRN-LEA) is proposed. Initially, the extracted images from the multimedia are collected from MSCOCO2014 dataset. Collected images are preprocessed to resize the image using federated neural collaborative filter (FedNCF). Later, resized images are given to feature extraction; morphological features like shape, structure, colour, pattern, and size are extracted based on synchro spline-kernelled chirplet extracting transform (SSCET). Finally, the extracted features are fed to Dual Transformer Residual Network (DTRN) for effectively classify the art images. In, general Dual Transformer Residual Network classifier does not express adapting optimization strategies to determine optimal parameters to ensure accurate art images detection system. Hence, the proposed method examined utilizing performance metrics like accuracy, precision, fl-score, sensitivity, peak signal to noise ratio (PSNR), error rate, and Structural similarity index (SSIM). Proposed DMAC-DTRN-LEA method attains 97% higher accuracy and 0.7% low error rate analysed to the existing methods respectively.