New media art design based on fast visual segmentation and 3D image processing

Abstract

Acquiring innovative styles and compositions from intricate and heterogeneous artistic imagery has emerged as a pivotal research quandry within contemporary new media art image conception. In a concerted effort to adeptly distill the quintessence of artistic styles and elements embedded within these visuals, an innovative methodology is posited herein, underpinned by an enhanced U-net segmentation framework and harmoniously fused with the surface extraction image reconstruction algorithm. This meticulous amalgamation endeavors to attain accurate segmentation and tridimensional reconstruction of the artistry encapsulated in these images. Primarily, the imagery is meticulously partitioned, culminating in an output that artfully encapsulates the inherent artistic attributes. Subsequently, this segmentation outcome is adeptly reconstituted, bestowing form to a three-dimensional artistry model. Empirical validation substantiates the efficacy of this approach, with the method’s Mean Intersection over the Union (MIoU) parameter yielding an impressive score of 0.939 in segmentation performance. Moreover, the peak signal-to-noise ratio and structural similarity attain commendable zeniths of 38.16 and 0.9808, respectively, underscoring the excellence of the reconstruction process. The proposed methodology demonstrates its prowess in exacting segmentation and comprehensive reconstruction of semantic intricacies and nuanced features pervading the realm of artistic imagery. Consequently, this novel methodology augments artists’ capacity to discern diverse artistic paradigms and fabricate superlative new media art compositions of heightened caliber.