- 【Updated on May 12, 2025】 Integration of CiNii Dissertations and CiNii Books into CiNii Research
- Trial version of CiNii Research Knowledge Graph Search feature is available on CiNii Labs
- Suspension and deletion of data provided by Nikkei BP
- Regarding the recording of “Research Data” and “Evidence Data”
Image Inpainting by Recursive Estimation Using Neural Network and Wavelet Transformation
Description
This paper proposes image inpainting algorithms in which the pixel values of the regions to be inpainting are recursively estimated by using multi-layered Perceptron from the original input image and its wavelet transformation. Instead of forward estimation by using the deep neural network such as convolutional neural network (CNN), we use shallow neural network and the pixel values are recursively estimated. To improve the estimation quality, wavelet transformation is also used as the input of the neural network. The effectiveness of the proposed approach was experimentally confirmed by using the face databases.
