RGBD Synergetic Model for Image Enhancement in Animation Advertisements

RGBD Synergetic Model for Image Enhancement in Animation Advertisements

Xuechun Wang, Wei Jiang
Copyright: © 2024 |Pages: 17
DOI: 10.4018/IJIIT.342478
Article PDF Download
Open access articles are freely available for download

Abstract

This paper proposes a depth image symbiosis model to solve the problem of insufficient depth image quality in animated advertising. The model uses image surface information and image edge cues as the main guidance information to obtain image symbiosis information. Research data show that the model designed in this paper performs well in convergence, and enters a stable convergence process when the number of iterations is less than 5. Its PSNR data curve has the highest position and best performance, while a composite model structure has been adopted. Compared with the unitary model, the PSNR of this model reaches 41 dB when the number of iterations reaches 5, and the convergence effect of the three-step training is also better. Finally, in practical applications, the average PSNR value of the model mentioned in this article is the highest, 37.1 dB. From this comprehensive perspective, the depth image enhancement model in this study has better comprehensive performance and can provide better image enhancement effects for animation advertising depth images.
Article Preview
Top

Presently, research in the field of image enhancement is continuously expanding. A study by Xu et al. (2020) proposed new insights into image processing for 360 degree video and images. Due to the spherical viewing range of 360 degree videos and images themselves, they had a larger data scale compared to ordinary videos and images. Therefore, compression and high-quality restoration of image data has become a key technology. The research analyzed the visual attention model and the spherical features of 360 degree images and looked forward to the future development of this technology (Xu et al., 2020). Ngugi et al. (2021) applied image processing technology to plant disease detection. The research combined machine learning technology with image processing technology, mainly using RGB images. The quality of CNN structural images was enhanced, which is the basis for the automatic recognition and formation of disease images. Lv et al. (2021) mainly used a CNN model with multiple branches to perform image enhancement on low light level images. This model combined attention mechanisms to form an end-to-end attention guidance strategy. At the same time, the form of dual attention map was adopted to achieve image enhancement and denoising. The first attention map could distinguish between underexposed and well-lit portions of the image, and the second attention map could distinguish between real texture and noise information. This method had strong fidelity.

Song et al. (2020) proposed an image enhancement model for underwater images. The model mainly depended on two optical parameters: background light and the transmission diagram. The model first performed a priori analysis of the dark channel of the underwater image and then used scene depth and reverse saturation to achieve color and contrast improvement of the image. At the same time, to balance the color and contrast of the image, white balance was used as the main post processing method. Its design method had excellent overall performance. Zhou et al. (2020) analyzed an image processing system that combines differentiators with traditional imaging systems. This image processing method could perform low-power and high-efficiency optical simulation, laying the foundation for further adaptive applications of computer vision processing and image enhancement processing technology.

On the other hand, as a relatively mature deep learning algorithm, the application of CNN in various fields is gradually deepening. Blanchet et al. (2020) proposed a stock price change prediction model combining CNN and Long Short-Term Memory (LSTM). This model could predict the closing price of stocks on the next day through the extracted characteristic data of stock price changes and could capture the closing price of stocks at different time points. This model had the ability to capture information characteristics over time, making it very suitable for predicting stock prices (Blanchet et al., 2020). Polsinelli et al. (2020) designed an automated diagnostic model for chest tomography images of COVID-19 based on light CNN. This model could effectively distinguish the chest CT images of COVID-19 from the healthy chest CT images. The model designed in this study could perform functions with more advantageous efficiency on notebook computers without hardware acceleration and had performance advantages (Polsinelli et al., 2020). Wang et al. (2020) designed an automated learning aid tool for CNN. This tool worked closely with CNN for users to use it more easily. Meanwhile, smooth transitions across abstraction levels were utilized to enhance the relevance between low-level mathematical computation and high-level model construction. This model could effectively assist users in algorithm learning.

Complete Article List

Search this Journal:
Reset
Volume 20: 1 Issue (2024)
Volume 19: 1 Issue (2023)
Volume 18: 4 Issues (2022): 3 Released, 1 Forthcoming
Volume 17: 4 Issues (2021)
Volume 16: 4 Issues (2020)
Volume 15: 4 Issues (2019)
Volume 14: 4 Issues (2018)
Volume 13: 4 Issues (2017)
Volume 12: 4 Issues (2016)
Volume 11: 4 Issues (2015)
Volume 10: 4 Issues (2014)
Volume 9: 4 Issues (2013)
Volume 8: 4 Issues (2012)
Volume 7: 4 Issues (2011)
Volume 6: 4 Issues (2010)
Volume 5: 4 Issues (2009)
Volume 4: 4 Issues (2008)
Volume 3: 4 Issues (2007)
Volume 2: 4 Issues (2006)
Volume 1: 4 Issues (2005)
View Complete Journal Contents Listing