Article Preview
TopIntroduction
With the rapid development of the multimedia field, art design is organically combined with multimedia, and the advent of a new multimedia era also brings many opportunities for art design (Coral et al., 2022). Digging into the constant changes of people’s aesthetic views, the visual impact of art design in the multimedia field should also be constantly developed and changed, and richer content should be given to make the visual effect of art design appear more stylish. Usually, multimedia is a type of information carrier, and its main presentation form is binary. It is mostly used to obtain, disseminate, process, and record information, such as digital animation, video images, sounds, images, graphics, and culture. Based on the multimedia context, the content of information dissemination is more and more diverse; the audience and the scope of dissemination are more and more extensive; the channels, media, and platforms of information dissemination are increasingly rich; and the timeliness and interactivity are getting higher and higher (Jarodzka, 2021; Bayraktar et al., 2019). Virtual reality (VR) technology is the most concerned image processing method in the era of big data. It simulates the surrounding environment and uses sensor devices to project and map virtual things, actions, or environments. If people watch video images, they can switch any viewpoint. This feature satisfies the visual pursuit of realism and immersion, and it is the development direction of the new generation multimedia video system (Cheng et al., 2021). At the sender, a small amount of viewpoint information is transmitted, and at the receiver, more viewing viewpoints are obtained through virtual viewpoint rendering technology (Shen et al., 2021; Shin et al., 2021).
VR is an effective way to use deep learning technology to process and visually evaluate video images. Zhang et al. (2018) proposed a no-reference video quality evaluation method based on a convolutional neural network (CNN) and resampling weakly supervised learning. This method mainly transmits the frequency histograms of distorted images and videos to resample the training set. Yan et al. (2019) proposed a no-reference image quality evaluation method based on CNN multitask learning that includes two tasks: prediction of statistical features of natural scenes and prediction of quality scores. However, virtual viewpoint rendering distortion is a variety of geometric distortions, such as artifacts, holes, and distortions. These distortions are different from coding and compression distortion all over the whole image and are mainly reflected in the texture edge region. Therefore, the existing 2D video image quality evaluation methods are directly used to evaluate the virtual viewpoint distortion, and the results obtained do not conform to the subjective perception of human eyes.
Therefore, in this research, I artistically process video images based on multimedia data fusion and reconstruct multimedia video images with an improved phase correlation algorithm to help artists complete their artistic design faster and better. In addition, I propose the use of a viewpoint image quality evaluation algorithm based on multifeature fusion and the visual characteristics of human eyes. This algorithm provides a reference for the visual evaluation of artistic image information.