Article Preview
TopIntroduction
Image, as the most direct expression of art, penetrates life, and in the field of deep learning-based data processing, many problems can be seen as image transformation tasks. The technical results of image conversion can be used both in the creation of new forms of contemporary artwork and in many other areas of data processing. In the last few years, as deep learning research has become more advanced, image transformation models and corresponding algorithms have been rapidly developed, in particular, the emergence of the generative adversarial network (GAN) (Yi, et al., 2019) and its improvement. GAN and image transformation technologies are also used for addressing real-world issues, such as model prediction and parameter identification, high-resolution reconstruction of images, image semantic segmentation, image restoration, image depth estimation, etc. These results show us that generative adversarial networks, as well as image transformation techniques, can be used to solve problems, such as data processing and image transformation in different domains (Cai, et al., 2021).
The creation of classical oil paintings, with the help of single still images, is time-consuming and laborious, both in terms of collecting images and labeling. A multimedia database is a database that stores and manages a large number of multimedia objects, such as audio data, image data, video data, sequence data, and hypertext data. Thus, how to mine the effective data of multiple modalities, such as video, audio, and text contained in multimedia and extract the correlation features of the research work is one of the main research topics in the area of computer vision, currently. Multimedia data mining includes many aspects. An example is image feature mining, including similarity search feature extraction, data cube, and multidimensional mining, association mining and classification, and predictive analysis.
In multimedia data cube and multidimensional feature extraction method, multimedia data cube is a kind of data cube for storing multidimensional data that also implements abstract data structures for multi-dimensional integrated queries at different abstraction layers, which can well support online analytical processing operations and data mining of multiple knowledge at multiple levels (Tang, et al., 2020). The complexity of multimedia data makes the structure of the multimedia data cube more complex with each additional dimension, and its physical implementation is a greater challenge. In practical applications, concepts of higher abstraction levels are often used instead of precise color values to represent knowledge rules, and conceptual inheritance can be used to generalize the original attributes to simplify the multimedia data cube structure. In addition, for multi-valued attributes, such as multiple colors of an image, the ones with the highest number of pixels of that color can be selected to make the corresponding multimedia data cube greatly simplified, and the mechanism of online analysis and mining is used to make the system with multiple data mining methods (Adnan & Akbar, 2019).