Implicit Neural Representation (INR) is a powerful tool for modeling complex functions, which effectively represents images and enables efficient interpolation. However, existing INRs do not represent the original image in a lossless manner. In this p...
Implicit Neural Representation (INR) is a powerful tool for modeling complex functions, which effectively represents images and enables efficient interpolation. However, existing INRs do not represent the original image in a lossless manner. In this paper, we propose a novel approach to achieve lossless neural representation by estimating the bits of the image using bit planewise implicit neural representation. We decompose images into bit planes and utilize SIREN activations to model the relationship between bit coordinates and corresponding outputs. We demonstrate that our method successfully represents the original image in a lossless manner. Furthermore, it achieves high quality representation in shorter running time compared to existing INR models, leading to efficient image neural representation.