Yuan Yuan

Beibei Lin, Yeying Jin, Wending Yan et al.

Masked autoencoder (MAE) shows that severe augmentation during training produces robust representations for high-level tasks. This paper brings the MAE-like framework to nighttime image enhancement, demonstrating that severe augmentation during training produces strong network priors that are resilient to real-world night haze degradations. We propose a novel nighttime image dehazing method with self-prior learning. Our main novelty lies in the design of severe augmentation, which allows our model to learn robust priors. Unlike MAE that uses masking, we leverage two key challenging factors of nighttime images as augmentation: light effects and noise. During training, we intentionally degrade clear images by blending them with light effects as well as by adding noise, and subsequently restore the clear images. This enables our model to learn clear background priors. By increasing the noise values to approach as high as the pixel intensity values of the glow and light effect blended images, our augmentation becomes severe, resulting in stronger priors. While our self-prior learning is considerably effective in suppressing glow and revealing details of background scenes, in some cases, there are still some undesired artifacts that remain, particularly in the forms of over-suppression. To address these artifacts, we propose a self-refinement module based on the semi-supervised teacher-student framework. Our NightHaze, especially our MAE-like self-prior learning, shows that models trained with severe augmentation effectively improve the visibility of input haze images, approaching the clarity of clear nighttime images. Extensive experiments demonstrate that our NightHaze achieves state-of-the-art performance, outperforming existing nighttime image dehazing methods by a substantial margin of 15.5% for MUSIQ and 23.5% for ClipIQA.

Yuan Yuan, Gene Cheung, Pascal Frossard et al.

Modern media data such as 360 videos and light field (LF) images are typically captured in much higher dimensions than the observers' visual displays. To efficiently browse high-dimensional media over bandwidth-constrained networks, a navigational streaming model is considered: a client navigates the large media space by dictating a navigation path to a server, who in response transmits the corresponding pre-encoded media data units (MDU) to the client one-by-one in sequence. Intra-coding an MDU (I-MDU) would result in a large bitrate but I-MDU can be randomly accessed, while inter-coding an MDU (P-MDU) using another MDU as a predictor incurs a small coding cost but imposes an order where the predictor must be first transmitted and decoded. From a compression perspective, the technical challenge is: how to achieve coding gain via inter-coding of MDUs, while enabling adequate random access for satisfactory user navigation. To address this problem, we propose landmarks, a selection of key MDUs from the high-dimensional media. Using landmarks as predictors, nearby MDUs in local neighborhoods are intercoded, resulting in a predictive MDU structure with controlled coding cost. It means that any requested MDU can be decoded by at most transmitting a landmark and an inter-coded MDU, enabling navigational random access. To build a landmarked MDU structure, we employ tree-structured vector quantizer (TSVQ) to first optimize landmark locations, then iteratively add/remove inter-coded MDUs as refinements using a fast branch-and-bound technique. Taking interactive LF images and viewport adaptive 360 images as illustrative applications, and I-, P- and previously proposed merge frames to intra- and inter-code MDUs, we show experimentally that landmarked MDU structures can noticeably reduce the expected transmission cost compared with MDU structures without landmarks.

Chengze Wang, Yuan Yuan, Qi Wang

In this paper, we present iDVO (inertia-embedded deep visual odometry), a self-supervised learning based monocular visual odometry (VO) for road vehicles. When modelling the geometric consistency within adjacent frames, most deep VO methods ignore the temporal continuity of the camera pose, which results in a very severe jagged fluctuation in the velocity curves. With the observation that road vehicles tend to perform smooth dynamic characteristics in most of the time, we design the inertia loss function to describe the abnormal motion variation, which assists the model to learn the consecutiveness from long-term camera ego-motion. Based on the recurrent convolutional neural network (RCNN) architecture, our method implicitly models the dynamics of road vehicles and the temporal consecutiveness by the extended Long Short-Term Memory (LSTM) block. Furthermore, we develop the dynamic hard-edge mask to handle the non-consistency in fast camera motion by blocking the boundary part and which generates more efficiency in the whole non-consistency mask. The proposed method is evaluated on the KITTI dataset, and the results demonstrate state-of-the-art performance with respect to other monocular deep VO and SLAM approaches.

Yuan Yuan, Siyuan Liu, Jiawei Zhang et al.

We consider the single image super-resolution problem in a more general case that the low-/high-resolution pairs and the down-sampling process are unavailable. Different from traditional super-resolution formulation, the low-resolution input is further degraded by noises and blurring. This complicated setting makes supervised learning and accurate kernel estimation impossible. To solve this problem, we resort to unsupervised learning without paired data, inspired by the recent successful image-to-image translation applications. With generative adversarial networks (GAN) as the basic component, we propose a Cycle-in-Cycle network structure to tackle the problem within three steps. First, the noisy and blurry input is mapped to a noise-free low-resolution space. Then the intermediate image is up-sampled with a pre-trained deep model. Finally, we fine-tune the two modules in an end-to-end manner to get the high-resolution output. Experiments on NTIRE2018 datasets demonstrate that the proposed unsupervised method achieves comparable results as the state-of-the-art supervised models.

Yuan Yuan, Gene Cheung, Patrick Le Callet et al.

A depth image provides partial geometric information of a 3D scene, namely the shapes of physical objects as observed from a particular viewpoint. This information is important when synthesizing images of different virtual camera viewpoints via depth-image-based rendering (DIBR). It has been shown that depth images can be efficiently coded using contour-adaptive codecs that preserve edge sharpness, resulting in visually pleasing DIBR-synthesized images. However, contours are typically losslessly coded as side information (SI), which is expensive if the object shapes are complex. In this paper, we pursue a new paradigm in depth image coding for color-plus-depth representation of a 3D scene: we pro-actively simplify object shapes in a depth and color image pair to reduce depth coding cost, at a penalty of a slight increase in synthesized view distortion. Specifically, we first mathematically derive a distortion upper-bound proxy for 3DSwIM---a quality metric tailored for DIBR-synthesized images. This proxy reduces interdependency among pixel rows in a block to ease optimization. We then approximate object contours via a dynamic programming (DP) algorithm to optimally trade off coding cost of contours using arithmetic edge coding (AEC) with our proposed view synthesis distortion proxy. We modify the depth and color images according to the approximated object contours in an inter-view consistent manner. These are then coded respectively using a contour-adaptive image codec based on graph Fourier transform (GFT) for edge preservation and HEVC intra. Experimental results show that by maintaining sharp but simplified object contours during contour-adaptive coding, for the same visual quality of DIBR-synthesized virtual views, our proposal can reduce depth image coding rate by up to 22% compared to alternative coding strategies such as HEVC intra.