<?xml version="1.0"?>
<rss version="2.0"
     xmlns:dc="http://purl.org/dc/elements/1.1/"
     xmlns:dcterms="http://purl.org/dc/terms/"
     xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
<title>IPOL</title>
<link>http://www.ipol.im/feed/</link>
<atom:link href="http://www.ipol.im/feed/preprints.rss" rel="self" type="application/atom+xml"/>
<description>IPOL Preprints — Latest public preprints from IPOL.</description>
<item>
	<title>A Survey of Metrics Used for Image Colorization Assessment</title>
	
	<dc:creator>Nicolas Maignan,
Fabien Pierre,
Fr&#xE9;d&#xE9;ric Sur</dc:creator>
	
	
	  <guid>http://www.ipol.im/pub/pre/632/</guid>
	
	<link>http://www.ipol.im/pub/pre/632/</link>
	<pubDate>Wed, 04 Nov 2026 12:51:51 +0100</pubDate>
	<dcterms:modified>2026-11-04T11:51:51Z</dcterms:modified>
	<description>The objective evaluation of image and video colorization remains challenging. Metrics such as PSNR and SSIM have limited ability to capture perceptual quality and typical colorization artifacts. Furthermore, they require a reference color image, which is generally unavailable for the colorization task. Alternatively, referenceless evaluation metrics have been proposed, aiming to better reflect human perception, yet few studies have analyzed their effectiveness. In this paper, we provide a comprehensive review of the evaluation metrics employed by recent colorization methods. We introduce a framework to assess these metrics based on their sensitivity to synthetic transformations that simulate common colorization defects, such as color bleeding or non-semantic colors. This evaluation is conducted both in the static case of images and in the temporal dimension for video colorization.	Besides, we propose extending to color images two computationally lightweight perceptual metrics previously defined in the literature for grayscale images.</description>
</item>
<item>
	<title>MBO Scheme for Local Chan-Vese Segmentation</title>
	
	<dc:creator>Kevin Bui,
Adina Ciomaga</dc:creator>
	
	
	  <guid>http://www.ipol.im/pub/pre/598/</guid>
	
	<link>http://www.ipol.im/pub/pre/598/</link>
	<pubDate>Fri, 16 Oct 2026 10:41:48 +0200</pubDate>
	<dcterms:modified>2026-10-16T08:41:48Z</dcterms:modified>
	<description>Robust to intensity inhomogeneity, the local Chan-Vese (LCV) model extends the classical Chan-Vese (CV) image segmentation method by incorporating local statistical information around each pixel. Originally, the LCV model was solved using a finite difference scheme, following the approach used for the CV model. As an alternative to the finite difference scheme, a more efficient algorithm based on the Merriman-Bence-Osher (MBO) scheme was later developed for the CV model. In this paper, we derive a similar MBO-based algorithm to solve the LCV model and propose an efficient implementation. The algorithm is developed for both two-phase and multiphase segmentation, and an extension to color images is also discussed. To demonstrate the effectiveness of the proposed approach, we apply it to a variety of grayscale and color images, including medical and microscopy images.</description>
</item>
<item>
	<title>A Modified Structure of Local Adaptive Hysteresis Smoothing for Image Denoising</title>
	
	<dc:creator>Ali Sabet Nasab,
Reza P.R. Hasanzadeh,
Pooriya Takhtechian</dc:creator>
	
	
	  <guid>http://www.ipol.im/pub/pre/595/</guid>
	
	<link>http://www.ipol.im/pub/pre/595/</link>
	<pubDate>Mon, 28 Sep 2026 18:48:19 +0200</pubDate>
	<dcterms:modified>2026-09-28T16:48:19Z</dcterms:modified>
	<description>Conventionally, Hysteresis Smoothing (HS)-based techniques rely on hysteresis thresholds to eliminate noise. The most efficient of the HS techniques, namely LAHS, performs well in image denoising; however, its structure has limitations in removing noise while preserving details. In this paper, a new framework, called Modified LAHS (MLAHS), is proposed to overcome these problems by modifying the cursor-width strategy and the HS process. The proposed method is evaluated quantitatively and qualitatively against other noise-reduction methods. The results show that the proposed structure significantly improves the performance of LAHS and also outperforms several efficient and widely used noise-suppression techniques.</description>
</item>
<item>
	<title>Linear Super-Resolution Through Translational Motion</title>
	
	<dc:creator>R&#xE9;my Abergel,
Andr&#xE9;s Almansa,
Lionel Moisan,
Camille No&#xFB;s</dc:creator>
	
	
	  <guid>http://www.ipol.im/pub/pre/550/</guid>
	
	<link>http://www.ipol.im/pub/pre/550/</link>
	<pubDate>Tue, 28 Jul 2026 17:16:20 +0200</pubDate>
	<dcterms:modified>2026-07-28T15:16:20Z</dcterms:modified>
	<description>The problem of computing a single super-resolved image from multiple shifted images of the same scene has been addressed in several ways in the literature, with different levels of assumptions on the image to reconstruct, but a frequent core step is the solution of the linear system derived from the image formation model. Here, we propose a detailed description of a direct (non-iterative) algorithm that computes the least-squares solution of the super-resolution problem in the general case of possibly non-integer super-resolution factors. We also provide a detailed analysis of the reconstruction error, both in the image and Fourier domains. In order to handle potential outliers, we build a robust (L1) variant of the algorithm, and illustrate its efficiency on synthetic and real data. We finally show how this linear and well-controlled process can be supplemented with a simple (still linear) spectral enhancement filter in order to better reveal super-resolved details without creating artifacts.</description>
</item>
<item>
	<title>Photo Response Non-Uniformity Extraction and Detection</title>
	
	<dc:creator>Benjamin Loison,
Quentin Bammey,
Rafael Grompone von Gioi,
Pablo Mus&#xE9;,
Marina Gardella</dc:creator>
	
	
	  <guid>http://www.ipol.im/pub/pre/662/</guid>
	
	<link>http://www.ipol.im/pub/pre/662/</link>
	<pubDate>Wed, 08 Jul 2026 15:03:50 +0200</pubDate>
	<dcterms:modified>2026-07-09T22:06:57Z</dcterms:modified>
	<description>Identifying the specific device model or instance that captured a given image is crucial in various forensic applications. In this context, the Photo Response Non-Uniformity (PRNU) plays a central role. The PRNU is a component of the image noise caused by the manufacturing process of digital cameras. Each pixel sensor responds slightly differently to the same light stimulus, under- or over-estimating the incoming signal in a way that depends on subtle, fixed imperfections unique to that sensor. This pixel-level variation creates a characteristic pattern that acts as a fingerprint of the camera. Because this fingerprint is intrinsic to the physical sensor, it remains consistent across different images taken with the same device. Furthermore, even among cameras of the same model, which share identical hardware and image processing pipelines, the PRNU pattern is distinct. The ability to extract and compare these fingerprints enables the attribution of an image to a specific device. The goal of this article is to provide a unified and transparent implementation of PRNU-based source camera identification methods, including several state-of-the-art approaches from the literature. We aim to systematically compare these methods, with particular emphasis on reproducibility, clarity of implementation, and ease of experimentation. Our objective is to deepen the understanding of the underlying assumptions, design choices, and processing steps involved in PRNU estimation and matching.</description>
</item>
<item>
	<title>Fixed Pattern Noise Reduction: Optimization-Based and Temporal High Pass Filter Methods</title>
	
	<dc:creator>Hortensia Barral,
Pablo Arias,
Axel Davy</dc:creator>
	
	
	  <guid>http://www.ipol.im/pub/pre/533/</guid>
	
	<link>http://www.ipol.im/pub/pre/533/</link>
	<pubDate>Mon, 29 Jun 2026 19:32:44 +0200</pubDate>
	<dcterms:modified>2026-06-29T17:32:44Z</dcterms:modified>
	<description>Fixed pattern noise (FPN) is a temporally coherent noise present on video due to the non-uniform response of the sensors. It is a common problem for infrared videos and can degrade the quality of the observation. In this work we study and compare two types of FPN removal methods: temporal high pass filter and optimization-based methods. Both are recursive real-time methods that perform very few operations per pixel. We show that the temporal high pass filter can be obtained as a particular case of optimization-based methods. We implement and compare several variants that have been proposed in the literature. </description>
</item>
<item>
	<title>QMSANet: A Quaternion Multi-Scale Attention Network for Color Image Denoising</title>
	
	<dc:creator>Yi Liu,
Qi Xie,
Yu Guo,
Guoqing Chen,
Boying Wu,
Deyu Meng,
Jean-Michel Morel,
Qiyu Jin,
Michael Kwok-Po Ng</dc:creator>
	
	
	  <guid>http://www.ipol.im/pub/pre/607/</guid>
	
	<link>http://www.ipol.im/pub/pre/607/</link>
	<pubDate>Tue, 04 Nov 2025 14:38:41 +0100</pubDate>
	<dcterms:modified>2025-11-04T13:38:41Z</dcterms:modified>
	<description>Color image denoising is a critical task in computer vision, often hindered by the underutilization of inter-channel correlations, resulting in color distortion and loss of fine details. We
propose QMSANet, a Quaternion Multi-Scale Attention Network, to address these challenges
by leveraging quaternion operations for color image denoising. Operating in the quaternion domain, QMSANet preserves channel dependencies across all processing stages, enhancing noise
suppression and detail retention. The network comprises three innovative modules: the Quaternion Multi-Scale Sparse Block (QMSB) for extracting multi-scale features with sparsity enforcement, the Quaternion Stacked Enhancement Block (QSEB) for refining deep features through
inter-channel interactions, and the Lightweight Quaternion Attention Block (LQAB) for adaptively focusing on salient features with minimal computational overhead. These modules collectively mitigate color deviation, detail loss, and edge artifacts. Extensive experiments on
benchmark datasets demonstrate that QMSANet outperforms state-of-the-art denoising models
in both synthetic and real-world noisy conditions. Typically, a blind denoiser exhibits diminished performance in comparison to a non-blind denoiser. However, QMASNet-B, a blind
denoiser constructed based on our model, also surpasses most of the comparison models. At
sigma = 15, QMASNet and QMASNet-B achieve PSNR improvements of 0.53 dB and 0.50 dB,
respectively, compared to the state-of-the-art method on CBSD68. Visual comparisons further
highlight its ability to preserve structural details. QMSANet offers a balanced, efficient solution
for high-quality color image denoising, with significant potential for real-world applications.</description>
</item>
<item>
	<title>SiamTE: Siamese Trace Erasing for Camera Trace Extraction</title>
	
	<dc:creator>Marina Gardella</dc:creator>
	
	
	  <guid>http://www.ipol.im/pub/pre/558/</guid>
	
	<link>http://www.ipol.im/pub/pre/558/</link>
	<pubDate>Wed, 17 Jul 2024 15:19:29 +0200</pubDate>
	<dcterms:modified>2024-10-05T22:16:56Z</dcterms:modified>
	<description>The camera trace is a signal embedded in the image during
the image formation process which implicitly encodes information about the camera processing chain. In the article &amp;#x27;Camera Trace Erasing&amp;#x27;, Chen et al. propose a Siamese Trace Erasing method aiming at extracting those traces. In order to do so, the authors design a novel hybrid loss for network training. This hybrid loss is defined as a combination of three different losses: the embedded similarity loss, the truncated fidelity loss and the cross-identity loss. In this article we briefly explore the method and its results.

**This is an MLBriefs article, the source code has not been reviewed!**&amp;#x3C;br&amp;#x3E;
**The original source code is [[available here|https://github.com/ngchc/CameraTE]] (last checked 2024/10/06).**&amp;#x3C;br&amp;#x3E;
 </description>
</item>
<item>
	<title>Termite Retinex</title>
	
	<dc:creator>Gabriele Simone,
Mauro Fiorentini,
Alessandro Rizzi</dc:creator>
	
	
	  <guid>http://www.ipol.im/pub/pre/531/</guid>
	
	<link>http://www.ipol.im/pub/pre/531/</link>
	<pubDate>Sun, 31 Mar 2024 20:13:30 +0200</pubDate>
	<dcterms:modified>2025-10-02T10:48:15Z</dcterms:modified>
	<description>The original presentation of Retinex, a spatial color correction and image enhancement 
algorithm modeling the Human Vision System, uses paths to explore the image in search of a local
reference white point. Here we present a spatial color algorithm, called Termite Retinex, 
with an alternative way to explore local properties of Retinex, replacing random paths with 
a colony of agents, which uses swarm intelligence to explore the image, determining in this way 
the locality of its filtering. We show the efficacy of Termite Retinex for unsupervised image enhancement,
dynamic range stretching and color correction for digital images.</description>
</item>
<item>
	<title>A Closed Form Solution to Natural Image Matting</title>
	
	<dc:creator>Mahdi Ranjbar,
Aissa Abdelaziz,
Mohammad Ali Jauhar</dc:creator>
	
	
	  <guid>http://www.ipol.im/pub/pre/532/</guid>
	
	<link>http://www.ipol.im/pub/pre/532/</link>
	<pubDate>Wed, 13 Mar 2024 19:46:46 +0100</pubDate>
	<dcterms:modified>2024-03-13T18:46:46Z</dcterms:modified>
	<description>Natural image matting refers to the process of estimating the foreground opacity, also known
as the alpha matte, of an input image and extracting the foreground layer. It is a fundamental
and challenging task in computer vision and finds applications in various fields related to image
processing such as film and image editing, advertising, and medical diagnosis. Numerous 
approaches have been proposed to address this challenge incorporating both traditional and deep
learning techniques. In this paper, we reproduce the paper &amp;#x22;A Closed Form Solution to Natural
Image Matting&amp;#x22;, also referred to as &amp;#x22;Closed Form Matting&amp;#x22;. The proposed algorithm is based
on local smoothness assumptions on foreground and background colors, the color-line model,
and provides a simple closed-form solution that requires sparse scribble constraints instead of
more cumbersome and hard-to-develop trimap constraints. We present extended derivations, a
Python-based implementation, an online demo, along with comparison on contemporary methods 
and ablation study on the effects of hyper-parameters.</description>
</item>
<item>
	<title>Experimental Improvements of Global Optimization Algorithms for Lipschitz Functions</title>
	
	<dc:creator>Perceval Beja-Battais,
Ga&#xEB;tan Serr&#xE9;,
Sophia Chirrane</dc:creator>
	
	
	  <guid>http://www.ipol.im/pub/pre/469/</guid>
	
	<link>http://www.ipol.im/pub/pre/469/</link>
	<pubDate>Mon, 05 Jun 2023 20:33:05 +0200</pubDate>
	<dcterms:modified>2023-06-05T18:33:05Z</dcterms:modified>
	<description>In this paper, we define an experimental context in which we tested the performances of LIPO and AdaLIPO, two global optimization algorithms for Lipschitz functions, introduced in [C. Malherbe and N. Vayatis, Global optimization of lipschitz functions, 2017]. We provide experimental proofs of the efficiency of
those algorithms, led numerical statistical analysis of our results, and suggested
two intuitive improvements from the vanilla version of the algorithms, referred
as LIPO-E and AdaLIPO-E. Within our test bench, these improvements allow
the algorithms to converge significantly faster and whenever they struggle to
find a better maximizer. Finally, we defined the scope of application of LIPO
and AdaLIPO. We show that they are very prone to the curse of dimensionality
and tend quickly to Pure Random Search when the dimension increases. We
provide source code for LIPO, AdaLIPO, and our enhanced versions</description>
</item>
<item>
	<title>An implementation of &amp;#x27;Efficient Multi-Stage Video Denoising Method&amp;#x27; and some variants</title>
	
	<dc:creator>Zhe Zheng,
Gabriele Facciolo,
Pablo Arias</dc:creator>
	
	
	  <guid>http://www.ipol.im/pub/pre/464/</guid>
	
	<link>http://www.ipol.im/pub/pre/464/</link>
	<pubDate>Tue, 28 Feb 2023 19:44:12 +0100</pubDate>
	<dcterms:modified>2023-03-14T12:42:14Z</dcterms:modified>
	<description>Recently, the field of image and video denoising has undergone a revolution
thanks to deep learning approaches. These methods outperform traditional 
model-based approaches in almost every image/video restoration problem. In this paper,
we propose an implementation of a recent approach proposed for video denoising,
namely Efficient Multi-stage Video Denoising method (EMVD). The method
has a lightweight and interpretable architecture consisting of three stages: temporal
fusion, denoising, and refinement stages. We reproduce this method and propose
three modifications aimed at improving its performance. (1) We apply motion
compensation to make better use of temporal redundancy, (2) we apply variance
stabilization to help this lightweight network deal with signal-dependent noise and
(3) we decouple occlusion detection and fusion weights prediction. We evaluate the
original method and the proposed modifications on a task of raw video denoising.</description>
</item>
<item>
	<title>Fixed Pattern Noise Reduction: Temporal High Pass Filter</title>
	
	<dc:creator>Arnaud Barral</dc:creator>
	
	
	  <guid>http://www.ipol.im/pub/pre/436/</guid>
	
	<link>http://www.ipol.im/pub/pre/436/</link>
	<pubDate>Fri, 25 Nov 2022 09:21:14 +0100</pubDate>
	<dcterms:modified>2022-11-25T08:21:14Z</dcterms:modified>
	<description>Temporal high pass filter methods are a family of methods for Fixed Pattern Noise (FPN) reduction. They are recursive real time methods that apply a high-pass temporal filter to remove the FPN. FPN is a temporally coherent noise present on video due to the non-uniformity response of the sensors. It is a common problem for infrared videos and can degrade the quality of the observation. In this work we will study and compare three classical temporal high pass filter FPNR methods.</description>
</item>

</channel>
</rss>
