Fabrizio Guerrini

Learning-based image coding is showing improved compression efficiency, while also offering a novel advantage in enabling computer vision tasks directly within the compressed domain. The latent representation created by deep learning methods inherently contains all visual features, without a computationally expensive synthesis process at the decoder. This paper is an invited extension of a previous solution for JPEG AI compressed domain face detection that adapts a RetinaFace-based detector to operate directly on the latent tensor. In addition to a former single-scale bridging solution, this work provides a novel multi-scale bridging architecture to enable a more effective multi-scale compressed domain face detection. The results show a significant performance gain, improving accuracy up to 20% for detection of tiny faces on the WIDER FACE dataset compared to single-scale bridging, and further narrowing the gap when compared to detection on uncompressed or JPEG AI decoded images. Furthermore, since the computationally expensive decoding step is bypassed and since the bridges consist of lower-complexity networks, the overall processing cost is significantly reduced. Single and multi-scale bridging, respectively, have about 10% and 32% the complexity of applying pixel domain face detection on decoded images. The proposed architecture is expected to be extended to other multiscale sensitive vision tasks, as JPEG AI is not specifically designed for any single downstream application.

Modern compression systems use linear transformations in their encoding and decoding processes, with transforms providing compact signal representations. While multiple data-dependent transforms for image/video coding can adapt to diverse statistical characteristics, assembling large datasets to learn each transform is challenging. Also, the resulting transforms typically lack fast implementation, leading to significant computational costs. Thus, despite many papers proposing new transform families, the most recent compression standards predominantly use traditional separable sinusoidal transforms. This paper proposes integrating a new family of Symmetry-based Graph Fourier Transforms (SBGFTs) of variable sizes into a coding framework, focusing on the extension from our previously introduced 8x8 SBGFTs to the general case of NxN grids. SBGFTs are non-separable transforms that achieve sparse signal representation while maintaining low computational complexity thanks to their symmetry properties. Their design is based on our proposed algorithm, which generates symmetric graphs on the grid by adding specific symmetrical connections between nodes and does not require any data-dependent adaptation. Furthermore, for video intra-frame coding, we exploit the correlations between optimal graphs and prediction modes to reduce the cardinality of the transform sets, thus proposing a low-complexity framework. Experiments show that SBGFTs outperform the primary transforms integrated in the explicit Multiple Transform Selection (MTS) used in the latest VVC intra-coding, providing a bit rate saving percentage of 6.23%, with only a marginal increase in average complexity. A MATLAB implementation of the proposed algorithm is available online at [1].

Tendons and ligaments play an important role to ensure mobility and stability. To correctly understand the characteristics of these fibrous collagenous connective tissues, it is fundamental to highlight their 3D microstructure. In this study a microtomography (microCT) system was used to acquire human hamstring tendons after performing specific preparations to enhance image contrast. Specifically, samples were treated either through chemical dehydration or by 2% of phosphotungstic acid ([ ]) in water ([ ]) or in 70% ethanol ([ ]) solution. Acquired images were elaborated using dedicated techniques based on 3D Hessian multiscale filter so as to highlight the fibrous structure and identify specific geometric features. For any strategy of sample preparation, the proposed approach resulted to be adequate for identifying fascicle features, thus obtaining structures with diameter in the range of 100–600 [ ] and proper longitudinal alignment. In conclusion, a novel contrast enhancement microCT protocol was designed and preliminarily validated for the microstructural analysis of fibrous tissues.

In the last years, due to the availability and easy of use of image editing tools, a large amount of fake and altered images have been produced and spread through the media and the Web. A lot of different approaches have been proposed in order to assess the authenticity of an image and in some cases to localize the altered (forged) areas. In this paper, we conduct a survey of some of the most recent image forgery detection methods that are specifically designed upon Deep Learning (DL) techniques, focusing on commonly found copy-move and splicing attacks. DeepFake generated content is also addressed insofar as its application is aimed at images, achieving the same effect as splicing. This survey is especially timely because deep learning powered techniques appear to be the most relevant right now, since they give the best overall performances on the available benchmark datasets. We discuss the key-aspects of these methods, while also describing the datasets on which they are trained and validated. We also discuss and compare (where possible) their performance. Building upon this analysis, we conclude by addressing possible future research trends and directions, in both deep learning architectural and evaluation approaches, and dataset building for easy methods comparison.

This paper explains how to describe any finite-energy signal through a unique representation consisting of an ordered set of positions and a sparse set of signals. This is obtained by designing an iterative decomposition through a series of mirror operations around those positions. The purpose is to find at any step of the decomposition the location that provides for the maximum decoupling between the even and odd components of the signal with respect to it. By limiting such even and odd components on three separate information bearing support, the algorithm can be iterated at infinity determining a sequence of positions. The per location information determines the optimal energy decoupling strategy at each stage providing remarkable sparsity in the representation. The resulting transformation leads to a 1-to-1 mapping. The approach is easily extended to finite-energy sequences, and in particular for sequences of finite length $N$ , at most $N$ iterations of the decomposition are required. Thanks to the sparsity of the resulting representation, experimental simulations demonstrate superior approximation capabilities of this proposed non-linear Mirror Transform with potential application in many domains such as approximation and coding. Its implementation has been made publicly available.