Harnessing Meta-Learning for Improving Full-Frame Video Stabilization

Video stabilization is a longstanding computer vision problem, particularly pixel-level synthesis solutions for video stabilization which synthesize full frames add to the complexity of this task. These techniques aim to stabilize videos by synthesizing full frames while enhancing the stability of the considered video. This intensifies the complexity of the task due to the distinct mix of unique motion profiles and visual content present in each video sequence, making robust generalization with fixed parameters difficult. In our study, we introduce a novel approach to enhance the performance of pixel-level synthesis solutions for video stabilization by adapting these models to individual input video sequences. The proposed adaptation exploits low-level visual cues accessible during test-time to improve both the stability and quality of resulting videos. We highlight the efficacy of our methodology of "test-time adaptation" through simple fine-tuning of one of these models, followed by significant stability gain via the integration of meta-learning techniques. Notably, significant improvement is achieved with only a single adaptation step. The versatility of the proposed algorithm is demonstrated by consistently improving the performance of various pixel-level synthesis models for video stabilization in real-world scenarios.

Efficient Model Agnostic Approach for Implicit Neural Representation Based Arbitrary-Scale Image Super-Resolution

Single image super-resolution (SISR) has experienced significant advancements, primarily driven by deep convolutional networks. Traditional networks, however, are limited to upscaling images to a fixed scale, leading to the utilization of implicit neural functions for generating arbitrarily scaled images. Nevertheless, these methodologies have imposed substantial computational demands as they involve querying every target pixel to a single resource-intensive decoder. In this paper, we introduce a novel and efficient framework, the Mixture of Experts Implicit Super-Resolution (MoEISR), which enables super-resolution at arbitrary scales with significantly increased computational efficiency without sacrificing reconstruction quality. MoEISR dynamically allocates the most suitable decoding expert to each pixel using a lightweight mapper module, allowing experts with varying capacities to reconstruct pixels across regions with diverse complexities. Our experiments demonstrate that MoEISR successfully reduces up to 73% in floating point operations (FLOPs) while delivering comparable or superior peak signal-to-noise ratio (PSNR).

NoiseTransfer: Image Noise Generation with Contrastive Embeddings

Deep image denoising networks have achieved impressive success with the help of a considerably large number of synthetic train datasets. However, real-world denoising is a still challenging problem due to the dissimilarity between distributions of real and synthetic noisy datasets. Although several real-world noisy datasets have been presented, the number of train datasets (i.e., pairs of clean and real noisy images) is limited, and acquiring more real noise datasets is laborious and expensive. To mitigate this problem, numerous attempts to simulate real noise models using generative models have been studied. Nevertheless, previous works had to train multiple networks to handle multiple different noise distributions. By contrast, we propose a new generative model that can synthesize noisy images with multiple different noise distributions. Specifically, we adopt recent contrastive learning to learn distinguishable latent features of the noise. Moreover, our model can generate new noisy images by transferring the noise characteristics solely from a single reference noisy image. We demonstrate the accuracy and the effectiveness of our noise model for both known and unknown noise removal.

Learning Task Agnostic Temporal Consistency Correction

Due to the scarcity of video processing methodologies, image processing operations are naively extended to the video domain by processing each frame independently. This disregard for the temporal connection in video processing often leads to severe temporal inconsistencies. State-of-the-art techniques that address these inconsistencies rely on the availability of unprocessed videos to siphon consistent video dynamics to restore the temporal consistency of frame-wise processed videos. We propose a novel general framework for this task that learns to infer consistent motion dynamics from inconsistent videos to mitigate the temporal flicker while preserving the perceptual quality for both the temporally neighboring and relatively distant frames. The proposed framework produces state-of-the-art results on two large-scale datasets, DAVIS and videvo.net, processed by numerous image processing tasks in a feed-forward manner. The code and the trained models will be released upon acceptance.

Rich CNN-Transformer Feature Aggregation Networks for Super-Resolution

Recent vision transformers along with self-attention have achieved promising results on various computer vision tasks. In particular, a pure transformer-based image restoration architecture surpasses the existing CNN-based methods using multi-task pre-training with a large number of trainable parameters. In this paper, we introduce an effective hybrid architecture for super-resolution (SR) tasks, which leverages local features from CNNs and long-range dependencies captured by transformers to further improve the SR results. Specifically, our architecture comprises of transformer and convolution branches, and we substantially elevate the performance by mutually fusing two branches to complement each representation. Furthermore, we propose a cross-scale token attention module, which allows the transformer to efficiently exploit the informative relationships among tokens across different scales. Our proposed method achieves state-of-the-art SR results on numerous benchmark datasets.

Restore from Restored: Single-image Inpainting

Recent image inpainting methods have shown promising results due to the power of deep learning, which can explore external information available from the large training dataset. However, many state-of-the-art inpainting networks are still limited in exploiting internal information available in the given input image at test time. To mitigate this problem, we present a novel and efficient self-supervised fine-tuning algorithm that can adapt the parameters of fully pre-trained inpainting networks without using ground-truth target images. We update the parameters of the pre-trained state-of-the-art inpainting networks by utilizing existing self-similar patches (i.e., self-exemplars) within the given input image without changing the network architecture and improve the inpainting quality by a large margin. Qualitative and quantitative experimental results demonstrate the superiority of the proposed algorithm, and we achieve state-of-the-art inpainting results on publicly available benchmark datasets.

Self-Supervised Adaptation for Video Super-Resolution

Recent single-image super-resolution (SISR) networks, which can adapt their network parameters to specific input images, have shown promising results by exploiting the information available within the input data as well as large external datasets. However, the extension of these self-supervised SISR approaches to video handling has yet to be studied. Thus, we present a new learning algorithm that allows conventional video super-resolution (VSR) networks to adapt their parameters to test video frames without using the ground-truth datasets. By utilizing many self-similar patches across space and time, we improve the performance of fully pre-trained VSR networks and produce temporally consistent video frames. Moreover, we present a test-time knowledge distillation technique that accelerates the adaptation speed with less hardware resources. In our experiments, we demonstrate that our novel learning algorithm can fine-tune state-of-the-art VSR networks and substantially elevate performance on numerous benchmark datasets.

Image Restoration by Solving IVP

Recent research on image restoration have achieved great success with the aid of deep learning technologies, but, many of them are limited to dealing SR with realistic settings. To alleviate this problem, we introduce a new formulation for image super-resolution to solve arbitrary scale image super-resolution methods. Based on the proposed new SR formulation, we can not only super-resolve images with multiple scales, but also find a new way to analyze the performance of super-resolving process. We demonstrate that the proposed method can generate high-quality images unlike conventional SR methods.

Learning Deep Video Stabilization without Optical Flow

Learning the necessary high-level reasoning for video stabilization without the help of optical flow has proved to be one of the most challenging tasks in the field of computer vision. In this work, we present an iterative frame interpolation strategy to generate a novel dataset that is diverse enough to formulate video stabilization as a supervised learning problem unassisted by optical flow. A major benefit of treating video stabilization as a pure RGB based generative task over the conventional optical flow assisted approaches is the preservation of content and resolution, which is usually obstructed in the latter approaches. To do so, we provide a new video stabilization dataset and train an efficient network that can produce competitive stabilization results in a fraction of the time taken to do the same with the recent iterative frame interpolation schema. Our method provides qualitatively and quantitatively better results than those generated through state-of-the-art video stabilization methods. To the best of our knowledge, this is the only work that demonstrates the importance of perspective in formulating video stabilization as a deep learning problem instead of replacing it with an inter-frame motion measure

Scene-Adaptive Video Frame Interpolation via Meta-Learning

Video frame interpolation is a challenging problem because there are different scenarios for each video depending on the variety of foreground and background motion, frame rate, and occlusion. It is therefore difficult for a single network with fixed parameters to generalize across different videos. Ideally, one could have a different network for each scenario, but this is computationally infeasible for practical applications. In this work, we propose to adapt the model to each video by making use of additional information that is readily available at test time and yet has not been exploited in previous works. We first show the benefits of `test-time adaptation' through simple fine-tuning of a network, then we greatly improve its efficiency by incorporating meta-learning. We obtain significant performance gains with only a single gradient update without any additional parameters. Finally, we show that our meta-learning framework can be easily employed to any video frame interpolation network and can consistently improve its performance on multiple benchmark datasets.