Natural Image Investigation Using Masked Image Encoding for Text Editing

Recognising and updating text is a typical photographic technique for changing text while maintaining the image's integrity. It offers several advantages, including the ability to rectify errors and restore text after an image has been damaged. Because the background is easily distorted, this is a difficult task. Masked autoencoders are self-supervised computer vision learners that are scalable. It works on the basis of two simple ideas: random parts of the original picture are masked, and missing pixels are restored. This study demonstrates how to alter the look of visual text down to the letter and digit level while maintaining the background's authenticity. Based on a masked autoencoder, LDN is built to update text on the natural images while maintaining the background texture. When it comes to assessing the 126 photographs of streets, posters, retail malls, and school settings, LDN performed brilliantly. It may effectively keep the background texture when modifying capital, lowercase, and numbers.