| Title |
Design of a CNN Model for Quality Assessment of Holographic Recording Results |
| Authors |
고철영(Cheolyoung Go) ; 황이환(Leehwan Hwang) |
| DOI |
https://doi.org/10.5573/ieie.2024.61.10.90 |
| Keywords |
Deep learning; Convolutional neural network; Hologram quality assessment |
| Abstract |
Among various deep learning technologies, CNN (Convolutional Neural Network) technique specialized in image processing is a convolutional neural network and is a neural network algorithm required for the image classification and quality evaluation tasks required in this paper. Holography is a technology that records and reproduces three-dimensional stereoscopic information on a two-dimensional flat surface of a recording medium using a recording light source. Since the holographic recording results are in an analog form, the process of evaluating the recorded results has been limited by mathematical calculation methods so far, or has been dependent on subjective and visual reviews. To solve these problems, a system that automatically evaluates the quality of holographic recording results quantitatively and objectively was implemented. We created 1,200 new images to be used in the learning dataset by classifying them according to grade. The images were classified into 6 grades so that they could be distinguished as overexposed or underexposed by the recording light source in the quality evaluation of the hologram. Since the captured images of holographic recording results are different from those of general digital photographs, a CNN model was designed to meet the criteria in order to distinguish them by differentiating the quality evaluation criteria. |