The importance of quantifying uncertainty in deep networks has become paramount for reliable real-world applications. In this paper, we propose a method to improve uncertainty estimation in medical Image-to-Image (I2I) translation. Our model integrates aleatoric uncertainty and employs Uncertainty-Aware Regularization (UAR) inspired by simple priors to refine uncertainty estimates and enhance reconstruction quality. We show that by leveraging simple priors on parameters, our approach captures more robust uncertainty maps, effectively refining them to indicate precisely where the network encounters difficulties, while being less affected by noise. Our experiments demonstrate that UAR not only improves translation performance, but also provides better uncertainty estimations, particularly in the presence of noise and artifacts. We validate our approach using two medical imaging datasets, showcasing its effectiveness in maintaining high confidence in familiar regions while accurately identifying areas of uncertainty in novel/ambiguous scenarios.

When characterising a digital camera spectrally or colourimetrically, the camera response to a generally diffusely reflecting colour chart is often employed. The recorded responses to the light incident from each colour patch are typically not linearly related to the power of the irradiance on the chart, and the irradiance varies with position on the chart. This necessitates a linearisation of the responses. We present a new single image colour chart-based estimation method of responses, that are linearly related to camera response values known as ground truth. The method estimates the spatial geometry of the irradiance incident on the chart attenuated by lens vignetting and compensates individually for volumetric and per colour channel non-linearities, including compensation for physical scene and camera properties in a pipeline of successive signal transformations between the estimated linear and the given recorded responses. The estimation is controlled by introducing a novel Additivity Principle of linear responses, which is derived from the spectral reflectances of the coloured surfaces on the colour chart, observing that linear relations of the spectral reflectances are equal to the relations of the corresponding linear responses. Crucially, the additivity principle is not subject to metamerism. The method is fundamentally solely reliant on a one-shot set of one triplet of response values sampled from each patch of a colour chart with known spectral reflectances, where rendition level, gray scale, illuminant, camera sensor curves, irradiance geometry, vignetting, moderate specular reflection, colour space, colour correction, gamut correction and noise level are unknown.

3D shape reconstruction from images is an active topic in computer vision. Shape-from-Shading is an important approach which requires the surface properties and light source position to infer the 3D shape. A L2 regularizer is typically used to penalize the irradiance equation. In this article, anisotropic diffusion (AD) is introduced as a regularizer to solve the image irradiance equation. The method is then compared with L1 and L2 regularization methods, where all of the three techniques are formulated using gradient descent. Results shows that with AD, edges can be better preserved. AD shows lower depth error and higher correlation when compared with L1 and L2 regularization methods.

3D shape reconstruction from images is an active topic in computer vision. Shape-from-Focus (SFF) is an important approach which requires image stack in a focus controlled manner to infer the 3D shape. In this article, 3D reconstruction of synthetic gastrointestinal regions is done using SFF. Image stack is generated in Blender software with focus controlled camera. A color focus measure is applied for shape recovery followed by a weighted L2 regularizer to estimate for inaccurate depth values. A precise comparison is done between recovered shape and ground truth data by measuring the depth error and correlation between them. Results shows that SFF technique will be practical for 3D reconstruction of GI regions with focus and motion controlled pillcams which is technologically feasible to implement.

Abstract Realistic renderings of materials with complex optical properties, such as goniochromatism and non-diffuse reflection, are difficult to achieve. In the context of the print and packaging industries, accurate visualisation of the complex appearance of such materials is a challenge, both for communication and quality control. In this paper, we characterise the bidirectional reflectance of two homogeneous print samples displaying complex optical properties. We demonstrate that in-plane retro-reflective measurements from a single input photograph, along with genetic algorithm-based BRDF fitting, allow to estimate an optimal set of parameters for reflectance models, to use for rendering. While such a minimal set of measurements enables visually satisfactory renderings of the measured materials, we show that a few additional photographs lead to more accurate results, in particular, for samples with goniochromatic appearance.

We propose a novel pathology-sensitive deep learning model (PS-DeVCEM) for frame-level anomaly detection and multi-label classification of different colon diseases in video capsule endoscopy (VCE) data. Our proposed model is capable of coping with the key challenge of colon apparent heterogeneity caused by several types of diseases. Our model is driven by attention-based deep multiple instance learning and is trained end-to-end on weakly labeled data using video labels instead of detailed frame-by-frame annotation. This makes it a cost-effective approach for the analysis of large capsule video endoscopy repositories. Other advantages of our proposed model include its capability to localize gastrointestinal anomalies in the temporal domain within the video frames, and its generality, in the sense that abnormal frame detection is based on automatically derived image features. The spatial and temporal features are obtained through ResNet50 and residual Long short-term memory (residual LSTM) blocks, respectively. Additionally, the learned temporal attention module provides the importance of each frame to the final label prediction. Moreover, we developed a self-supervision method to maximize the distance between classes of pathologies. We demonstrate through qualitative and quantitative experiments that our proposed weakly supervised learning model gives a superior precision and F1-score reaching, 61.6% and 55.1%, as compared to three state-of-the-art video analysis methods respectively. We also show our model’s ability to temporally localize frames with pathologies, without frame annotation information during training. Furthermore, we collected and annotated the first and largest VCE dataset with only video labels. The dataset contains 455 short video segments with 28,304 frames and 14 classes of colorectal diseases and artifacts. Dataset and code supporting this publication will be made available on our home page.

This paper is an attempt to provide a rigorous basis to the interpolation of the MacAdam ellipses. It starts by defining criteria used to compare the different possible interpolations. Then several interpolation strategies are compared. The main conclusion that arises from this comparison is that the ellipses should not be interpolated based on the coefficients of the matrices of the corresponding scalar products, as MacAdam suggested, but on the coefficients of their inverses. It also appears that the $uv$ parameters tend to gives better results than the $xy$ and $ab$ parameters.

Aiming at understanding the role of short-term memory in subjective image quality assessment, we report and compare results from two pair-comparison methods: stimuli shown side-by-side versus stimuli shown one after the other. Our results suggest that there is a significant chance that an observer will make different quality assessments in the two setups.

Image reproduction devices, such as displays or printers, can reproduce only a limited set of colors, denoted the color gamut. The gamut depends on both theoretical and technical limitations. Reproduction device gamuts are significantly different from acquisition device gamuts. These facts raise the problem of reproducing similar color images across different devices. This is well known as the gamut mapping problem. Gamut mapping algorithms have been developed mainly using colorimetric pixel-wise principles, without considering the spatial properties of the image. The recently proposed multilevel gamut mapping approach takes spatial properties into account and has been demonstrated to outperform spatially invariant approaches. However, they have some important drawbacks. To analyze these drawbacks, we build a common framework that encompasses at least two important previous multilevel gamut mapping algorithms. Then, when the causes of the drawbacks are understood, we solve the typical problem of possible hue shifts. Next, we design appropriate operators and functions to strongly reduce both haloing and possible undesired over compression. We use challenging synthetic images, as well as real photographs, to practically show that the improvements give the expected results.

No abstract available.