Incorporating diffusion models in the image compression domain has the potential to produce realistic and detailed reconstructions, especially at extremely low bitrates. Previous methods focus on using diffusion models as expressive decoders robust to quantization errors in the conditioning signals, yet achieving competitive results in this manner requires costly training of the diffusion model and long inference times due to the iterative generative process. In this work we formulate the removal of quantization error as a denoising task, using diffusion to recover lost information in the transmitted image latent. Our approach allows us to perform less than 10\% of the full diffusion generative process and requires no architectural changes to the diffusion model, enabling the use of foundation models as a strong prior without additional fine tuning of the backbone. Our proposed codec outperforms previous methods in quantitative realism metrics, and we verify that our reconstructions are qualitatively preferred by end users, even when other methods use twice the bitrate.
Due to old CRT display technology and limited transmission bandwidth, early film and TV broadcasts commonly used interlaced scanning. This meant each field contained only half of the information. Since modern displays require full frames, this has spurred research into deinterlacing, i.e. restoring the missing information in legacy video content. In this paper, we present a deep-learning-based method for deinterlacing animated and live-action content. Our proposed method supports bidirectional spatio-temporal information propagation across multiple scales to leverage information in both space and time. More specifically, we design a Flow-guided Refinement Block (FRB) which performs feature refinement including alignment, fusion, and rectification. Additionally, our method can process multiple fields simultaneously, reducing per-frame processing time, and potentially enabling real-time processing. Our experimental results demonstrate that our proposed method achieves superior performance compared to existing methods.
Face recognition models embed a face image into a low-dimensional identity vector containing abstract encodings of identity-specific facial features that allow individuals to be distinguished from one another. We tackle the challenging task of inverting the latent space of pre-trained face recognition models without full model access (i.e. black-box setting). A variety of methods have been proposed in literature for this task, but they have serious shortcomings such as a lack of realistic outputs, long inference times, and strong requirements for the data set and accessibility of the face recognition model. Through an analysis of the black-box inversion problem, we show that the conditional diffusion model loss naturally emerges and that we can effectively sample from the inverse distribution even without an identity-specific loss. Our method, named identity denoising diffusion probabilistic model (ID3PM), leverages the stochastic nature of the denoising diffusion process to produce high-quality, identity-preserving face images with various backgrounds, lighting, poses, and expressions. We demonstrate state-of-the-art performance in terms of identity preservation and diversity both qualitatively and quantitatively. Our method is the first black-box face recognition model inversion method that offers intuitive control over the generation process and does not suffer from any of the common shortcomings from competing methods.
Noise synthesis is a challenging low-level vision task aiming to generate realistic noise given a clean image along with the camera settings. To this end, we propose an effective generative model which utilizes clean features as guidance followed by noise injections into the network. Specifically, our generator follows a UNet-like structure with skip connections but without downsampling and upsampling layers. Firstly, we extract deep features from a clean image as the guidance and concatenate a Gaussian noise map to the transition point between the encoder and decoder as the noise source. Secondly, we propose noise synthesis blocks in the decoder in each of which we inject Gaussian noise to model the noise characteristics. Thirdly, we propose to utilize an additional Style Loss and demonstrate that this allows better noise characteristics supervision in the generator. Through a number of new experiments, we evaluate the temporal variance and the spatial correlation of the generated noise which we hope can provide meaningful insights for future works. Finally, we show that our proposed approach outperforms existing methods for synthesizing camera noise.
Recently, significant progress has been made in learned image and video compression. In particular the usage of Generative Adversarial Networks has lead to impressive results in the low bit rate regime. However, the model size remains an important issue in current state-of-the-art proposals and existing solutions require significant computation effort on the decoding side. This limits their usage in realistic scenarios and the extension to video compression. In this paper, we demonstrate how to leverage knowledge distillation to obtain equally capable image decoders at a fraction of the original number of parameters. We investigate several aspects of our solution including sequence specialization with side information for image coding. Finally, we also show how to transfer the obtained benefits into the setting of video compression. Overall, this allows us to reduce the model size by a factor of 20 and to achieve 50% reduction in decoding time.
Image restoration has seen great progress in the last years thanks to the advances in deep neural networks. Most of these existing techniques are trained using full supervision with suitable image pairs to tackle a specific degradation. However, in a blind setting with unknown degradations this is not possible and a good prior remains crucial. Recently, neural network based approaches have been proposed to model such priors by leveraging either denoising autoencoders or the implicit regularization captured by the neural network structure itself. In contrast to this, we propose using normalizing flows to model the distribution of the target content and to use this as a prior in a maximum a posteriori (MAP) formulation. By expressing the MAP optimization process in the latent space through the learned bijective mapping, we are able to obtain solutions through gradient descent. To the best of our knowledge, this is the first work that explores normalizing flows as prior in image enhancement problems. Furthermore, we present experimental results for a number of different degradations on data sets varying in complexity and show competitive results when comparing with the deep image prior approach.
Deep learning based image compression has recently witnessed exciting progress and in some cases even managed to surpass transform coding based approaches that have been established and refined over many decades. However, state-of-the-art solutions for deep image compression typically employ autoencoders which map the input to a lower dimensional latent space and thus irreversibly discard information already before quantization. Due to that, they inherently limit the range of quality levels that can be covered. In contrast, traditional approaches in image compression allow for a larger range of quality levels. Interestingly, they employ an invertible transformation before performing the quantization step which explicitly discards information. Inspired by this, we propose a deep image compression method that is able to go from low bit-rates to near lossless quality by leveraging normalizing flows to learn a bijective mapping from the image space to a latent representation. In addition to this, we demonstrate further advantages unique to our solution, such as the ability to maintain constant quality results through re-encoding, even when performed multiple times. To the best of our knowledge, this is the first work to explore the opportunities for leveraging normalizing flows for lossy image compression.
The field of neural image compression has witnessed exciting progress as recently proposed architectures already surpass the established transform coding based approaches. While, so far, research has mainly focused on architecture and model improvements, in this work we explore content adaptive optimization. To this end, we introduce an iterative procedure which adapts the latent representation to the specific content we wish to compress while keeping the parameters of the network and the predictive model fixed. Our experiments show that this allows for an overall increase in rate-distortion performance, independently of the specific architecture used. Furthermore, we also evaluate this strategy in the context of adapting a pretrained network to other content that is different in visual appearance or resolution. Here, our experiments show that our adaptation strategy can largely close the gap as compared to models specifically trained for the given content while having the benefit that no additional data in the form of model parameter updates has to be transmitted.
Traditional approaches for color propagation in videos rely on some form of matching between consecutive video frames. Using appearance descriptors, colors are then propagated both spatially and temporally. These methods, however, are computationally expensive and do not take advantage of semantic information of the scene. In this work we propose a deep learning framework for color propagation that combines a local strategy, to propagate colors frame-by-frame ensuring temporal stability, and a global strategy, using semantics for color propagation within a longer range. Our evaluation shows the superiority of our strategy over existing video and image color propagation methods as well as neural photo-realistic style transfer approaches.
Minimization of regularized losses is a principled approach to weak supervision well-established in deep learning, in general. However, it is largely overlooked in semantic segmentation currently dominated by methods mimicking full supervision via "fake" fully-labeled training masks (proposals) generated from available partial input. To obtain such full masks the typical methods explicitly use standard regularization techniques for "shallow" segmentation, e.g. graph cuts or dense CRFs. In contrast, we integrate such standard regularizers directly into the loss functions over partial input. This approach simplifies weakly-supervised training by avoiding extra MRF/CRF inference steps or layers explicitly generating full masks, while improving both the quality and efficiency of training. This paper proposes and experimentally compares different losses integrating MRF/CRF regularization terms. We juxtapose our regularized losses with earlier proposal-generation methods using explicit regularization steps or layers. Our approach achieves state-of-the-art accuracy in semantic segmentation with near full-supervision quality.