This paper discusses a super-resolution (SR) system implemented on a mobile device. We utilized an Android device's camera to take successive shots and applied a classical multiple-image super-resolution (SR) technique that utilized a set of low-resolution (LR) images. Images taken from the mobile device are subjected to our proposed filtering scheme wherein images that have noticeable presence of blur are discarded to avoid outliers from affecting the produced high-resolution (HR) image. The remaining subset of images are subjected to non-local means denoising, then feature-matched against the first reference LR image. Successive images are then aligned with respect to the first image via affine and perspective warping transformations. The LR images are then upsampled using bicubic interpolation. An L 2 -norm minimization approach, which is essentially taking the pixel-wise mean of the aligned images, is performed to produce the final HR image. Our study shows that our proposed method performs better than the bicubic interpolation, which makes its implementation in a mobile device quite feasible. We have also proven in our experiments that there are substantial differences from images captured using burst mode that can be utilized by an SR algorithm to create an HR image.
In this work, we present a network architecture with parallel convolutional neural networks (CNN) for removing perspective distortion in images. While other works generate corrected images through the use of generative adversarial networks or encoder-decoder networks, we propose a method wherein three CNNs are trained in parallel, to predict a certain element pair in the 3×3 transformation matrix, M^. The corrected image is produced by transforming the distorted input image using M^−1. The networks are trained from our generated distorted image dataset using KITTI images. Experimental results show promise in this approach, as our method is capable of correcting perspective distortions on images and outperforms other state-of-the-art methods. Our method also recovers the intended scale and proportion of the image, which is not observed in other works.
In this study, we showcase a mobileaugmented reality application where a user places various 3D models in atabletop scene. The scene is captured and then rendered as Claude Monet’s impressionistic art style. One possibleuse case for this application is to demonstrate the behavior of the impressionistic art style of Claude Monet, byapplying this to tabletop scenes, which can be useful especially for art students. This allows the user to create theirown "still life" composition and study how the scene is painted. Our proposed framework is composed of threesteps. The system first identifies the context of the tabletop scene, through GIST descriptors, which are used asfeatures to identify the color palette to be used for painting. Our application supports three different color palettes,representing different eras of Monet’s work. The second step performs color mixing of two different colors in thechosen palette. The last step involves applying a three-stage brush stroke algorithm where the image is renderedwith a customized brush stroke pattern applied in each stage. While deep learning techniques are already capableof performing style transfer from paintings to real-world images, such as the success of CycleGAN, results showthat our proposed framework achieves comparable performance to deep learning style transfer methods on tabletopscenes.
In this study, we showcase a mobile augmented reality application where a user places various 3D models in a tabletop scene. The scene is captured and then rendered as Claude Monet's impressionistic art style. One possible use case for this application is to demonstrate the behavior of the impressionistic art style of Claude Monet, by applying this to tabletop scenes, which can be useful especially for art students. This allows the user to create their own "still life" composition and study how the scene is painted. Our proposed framework is composed of three steps. The system first identifies the context of the tabletop scene, through GIST descriptors, which are used as features to identify the color palette to be used for painting. Our application supports three different color palettes, representing different eras of Monet's work. The second step performs color mixing of two different colors in the chosen palette. The last step involves applying a three-stage brush stroke algorithm where the image is rendered with a customized brush stroke pattern applied in each stage. While deep learning techniques are already capable of performing style transfer from paintings to real-world images, such as the success of CycleGAN, results show that our proposed framework achieves comparable performance to deep learning style transfer methods on tabletop scenes.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.