Deep Spatial–focal Network for Depth from Focus

Traditional depth from focus (DFF) methods obtain image a set of differently focused color images. They detect in-focus region at each by measuring the sharpness observed textures. However, estimating arbitrary texture is not trivial task especially when there are limited or intensity variations in an image. Recent deep learning based DFF approaches have shown that collective estimation images on large body training samples outperforms traditional with challenging target objects textureless glaring surfaces. In this article, we propose spatial‐focal convolutional neural network encodes correlations between consecutive fed to order. way, our understands pattern blur changes pixel volumetric input three-dimensional space. Extensive quantitative and qualitative evaluations existing three public data sets show proposed method prior estimation.