State-of-the-art 2D image compression schemes rely on the power of convolutional neural networks (CNNs). Although CNNs offer promising perspectives for 2D image compression, extending such models to omnidirectional images is not straightforward. First, omnidirectional images have specific spatial and statistical properties that can not be fully captured by current CNN models. Second, basic mathematical operations composing a CNN architecture, e.g. , translation and sampling, are not well-defined on the sphere. In this paper, we study the learning of representation models for omnidirectional images and propose to use the properties of HEALPix uniform sampling of the sphere to redefine the mathematical tools used in deep learning models for omnidirectional images. In particular, we: i) propose the definition of a new convolution operation on the sphere that keeps the high expressiveness and the low complexity of a classical 2D convolution; ii) adapt standard CNN techniques such as stride, iterative aggregation, and pixel shuffling to the spherical domain; and then iii) apply our new framework to the task of omnidirectional image compression. Our experiments show that our proposed on-the-sphere solution leads to a better compression gain that can save 13.7% of the bit rate compared to similar learned models applied to equirectangular images. Also, compared to learning models based on graph convolutional networks, our solution supports more expressive filters that can preserve high frequencies and provide a better perceptual quality of the compressed images. Such results demonstrate the efficiency of the proposed framework, which opens new research venues for other omnidirectional vision tasks to be effectively implemented on the sphere manifold.

In this paper, we propose a new interactive compression scheme for omnidirectional images. This requires two characteristics: efficient compression of data, to lower the storage cost, and random access ability to extract part of the compressed stream requested by the user (for reducing the transmission rate). For efficient compression, data needs to be predicted by a series of references that have been pre-defined and compressed. This contrasts with the spirit of random accessibility. We propose a solution for this problem based on incremental codes implemented by rate-adaptive channel codes. This scheme encodes the image while adapting to any user request and leads to an efficient coding that is flexible in extracting data depending on the available information at the decoder. Therefore, only the information that is needed to be displayed at the user's side is transmitted during the user's request, as if the request was already known at the encoder. The experimental results demonstrate that our coder obtains a better transmission rate than the state-of-the-art tile-based methods at a small cost in storage. Moreover, the transmission rate grows gradually with the size of the request and avoids a staircase effect, which shows the perfect suitability of our coder for interactive transmission.

In this work, we study the restoration of low-light images with outdoor scenes without ground truth. Until now, approaches in the literature have avoided using the Retinex decomposition model in an unsupervised way or have added constraining priors on the searched components. We propose here to relax the constraint of a grayscale illumination of the Retinex model. Indeed, according to the physics of light, it should include a colored illumination. Resulting from this new decomposition model, we formulate a new deep learning-based architecture inspired by the style transfer methods. Our method enables us to visualize the illumination (i.e. a complex style with the same dimensions as an image) and the reflectance (i.e. the content). It achieves more visually pleasing components compared to the state-of-the-art i.e. without artifact, without noise amplification and without hallucination with a simple restoration for each of the components.

On-the-sphere compression of omnidirectional videos is a very promising approach. First, it saves computational complexity as it avoids to project the sphere onto a 2D map, as classically done. Second, and more importantly, it allows to achieve a better rate-distortion tradeoff, since neither the visual data nor its domain of definition are distorted. In this paper, the on-the-sphere compression [1] for omnidirectional still images is extended to videos. We first propose a complete review of existing spherical motion models. Then we pro-pose a new one called tangent-linear+t. We finally propose a rate-distortion optimized algorithm to locally choose the best motion model for efficient motion estimation/compensation. For that purpose, we additionally propose a finer search pattern, called spherical-uniform, for the motion parameters, which leads to a more accurate block prediction. The novel algorithm leads to rate-distortion gains compared to methods based on a unique motion model.

This paper introduces a novel approach called semantic boosting regression (SBR), leveraging the principles of boosting algorithms in symbolic regression using a Memetic Semantic GP for Symbolic Regression (MSGP) algorithm as weak learners. Memetic computation facilitates the integration of domain knowledge into a population-based approach, and semantic-based algorithms enhance local improvements to achieve targeted outputs. The fusion of memetic and semantic approaches allows us to augment the exploration and exploitation capabilities inherent in Genetic Programming (GP) and identify concise symbolic expressions that maintain interpretability without compromising the expressive power of symbolic regression. Our approach echoes the boosting algorithm’s characteristic, where weak learners (e.g., MSGP) are sequentially improved upon, focusing on correcting previous errors and continuously enhancing overall performance. This iterative strategy, intrinsic to boosting methods, is adeptly adapted to our SBR model. Experimental results demonstrate that our memetic-semantic approach has equal or better performance when compared to state-of-the-art evolutionary-based techniques when addressing real-world symbolic regression challenges. This advancement helps tackle the bloating issue in GP and significantly improves generalization capabilities. However, akin to classic boosting algorithms, one limitation of our approach is the increased computational cost due to the sequential training of boosting learners.