Contour Completion using Deep Structural Priors

TL;DR

This paper presents a novel deep structural prior framework that enables contour completion in images using only a single image for training, mimicking human perceptual abilities without requiring additional data.

Contribution

It introduces a new iterative model for contour completion that does not need prior knowledge of missing regions and trains on a single image, advancing unsupervised image inpainting techniques.

Findings

Successfully completes disconnected contours and fragmented lines.

Operates without prior knowledge of missing regions.

Trains on a single image without additional data.

Abstract

Humans can easily perceive illusory contours and complete missing forms in fragmented shapes. This work investigates whether such capability can arise in convolutional neural networks (CNNs) using deep structural priors computed directly from images. In this work, we present a framework that completes disconnected contours and connects fragmented lines and curves. In our framework, we propose a model that does not even need to know which regions of the contour are eliminated. We introduce an iterative process that completes an incomplete image and we propose novel measures that guide this to find regions it needs to complete. Our model trains on a single image and fills in the contours with no additional training data. Our work builds a robust framework to achieve contour completion using deep structural priors and extensively investigate how such a model could be implemented.