Phase-Aware Wavelet-Based-Scattering Encoder-Decoder for Dense Predictions
Researchers have developed a new Phase-Aware Scattering Encoder-Decoder model designed to improve dense prediction tasks in computer vision. This model enhances scattering transforms by preserving spatial structure and phase information, which are typically lost in global averaging. Initial tests on image denoising show significant improvements in PSNR, and a preliminary study on skin lesion segmentation is also underway. AI
IMPACT Introduces a novel deep learning architecture that could improve performance on image denoising and segmentation tasks.