This paper presents an approach for estimating H&E stain and phase information from microscopy images using encoder-decoder neural networks. We propose a two-stage framework to address the challenges of virtual staining and phase information generation from limited datasets. The first stage involves a Stain model that converts grayscale brightfield images into virtual H&E-stained images using a modified encoder-decoder network with 65 layers. The second stage utilizes a Phase model to transform virtual H&E-stained images into quantitative phase images, leveraging a 33-layer encoder-decoder network. Experimental results demonstrate the effectiveness of the proposed models on datasets of skin and breast tissue samples, achieving high structural similarity (SSIM) and peak signal-to-noise ratio (PSNR) metrics. This framework offers a promising solution for reducing reliance on physical staining processes, mitigating domain shift issues, and advancing the field of label-free microscopy imaging.