Geometry based interpolation improves image recovery accuracy and speed
Geometry-Aware Interpolation for Inverse Problems
Computer Vision and Pattern RecognitionMachine Learning
Summary
Fixing images that are damaged or missing parts is hard, especially when images have complex shapes. The paper presents a new method where instead of guessing pixel values from nearby fixed spots, the method looks at the natural geometric structure of the image data to fill in missing parts more smartly. This geometry-aware method helps the computer learn better and faster while creating clearer reconstructed images. The authors tested their method on image completion and recognition tasks with good results.
What this means in practice
- •For image processing teams: Improve image completion quality and speed in photo editing and restoration software.
- •For computer vision engineers: Enhance classification accuracy when dealing with incomplete or degraded visual data inputs.
Authors
Pourya Shamsolmoali, Masoumeh Zareapoor
Abstract
Recovering missing visual information from incomplete or degraded images remains a challenging problem, particularly when spatial variations and complex image structures affect reconstruction accuracy. Although Spatial Transformer Networks (STNs) facilitate spatial feature alignment, their reliance on conventional linear or bilinear interpolation limits their ability to capture the underlying geometric characteristics of image data. Furthermore, interpolation based on a restricted set of neighbouring pixels can introduce inconsistencies in gradient propagation, potentially affecting optimisation stability. To address these limitations, we propose a Geometry-Aware Interpolation Network (GAIN), a variational autoencoder-based framework that incorporates manifold geometry into spatial transformation and image reconstruction. Rather than estimating pixel intensities solely from fixed spatial neighbours, GAIN exploits the local tangent structure of the data manifold to construct geometry-informed intensity approximations. An entropy-based regularisation mechanism further guides the estimation of spatial transformation parameters, promoting more adaptive interpolation. In addition, a convolutional normalisation strategy is introduced to regulate the magnitude of non-zero activations and improve gradient stability during training. Extensive experiments on challenging image completion and classification benchmarks demonstrate the effectiveness of GAIN in improving predictive accuracy while reducing computational cost compared with existing methods.