Convolutional Neural Networks (CNNs) have become indispensable tools in skin cancer classification, aiding clinical experts to achieve earlier and more accurate diagnoses, improving treatment outcomes, and driving advancements in medical research. Despite their pivotal role, the most popular CNN architecture families exhibit a critical issue related to the distribution and quantity of available data, potentially compromising their performance and generalization abilities. This challenge is commonly overlooked in most skin lesion classification papers, which mainly rely on weighted classification techniques. Directly using appropriately dataset balancing or Transfer Learning (TL) methods, as suggested in recent studies, has the potential to deliver more satisfactory results, providing a more effective approach to addressing this issue. In the effort to tackle this problem, we provide a comprehensive quantitative evaluation aimed at identifying the most critical new emerging computational aspects and the related effective techniques. Specifically, we propose twelve Computational Models (CMs) based on four prominent CNN models with increasing structural complexity. We assess their effectiveness in both pretrained and unpretrained versions, incorporating TL strategies as well. Our experiments focus on the ISIC 2018 image dataset, a benchmark widely recognized for its extensive application in skin cancer research yet challenged by significant class imbalance issues. To mitigate this, we also randomly extracted a balanced image subset from ISIC 2018 for evaluation purposes. The experimental results, regarding four different scenarios, provide valuable insights into the design and utilization of CNNs for skin lesion classification, laying the groundwork for further investigations.

Morphological Characterization of figs leaves

Spatial color algorithms (SCAs) are algorithms grounded in the retinex theory of color sensation that, mimicking the human visual system, perform image enhancement based on the spatial arrangement of the scene. Despite their established role in image enhancement, their potential as dequantizers has never been investigated. Here, we aim to assess the effectiveness of SCAs in addressing the dual objectives of color dequantization and image enhancement at the same time. To this end, we propose the term dequantenhancement. In this paper, through two experiments on a dataset of images, SCAs are evaluated through two distinct pathways: first, quantization followed by filtering to assess both dequantization and enhancement; and second, filtering applied to original images before quantization as further investigation of mainly the dequantization effect. The results are presented both qualitatively, with visual examples, and quantitatively, through metrics including the number of colors, retinal-like subsampling contrast (RSC), and structural similarity index (SSIM).

Precision Viticulture (PV) is becoming an active and interdisciplinary research field since it requires solving interesting research issues to concretely answer the demands of specific use cases. A challenging problem in this context is the development of automatic methods for yield estimation. Computer vision methods can contribute to the accomplishment of this task, especially those that can replicate what winemakers do manually. In this paper, an automatic artificial intelligence method for grape bunch detection from RGB images is presented. A customized Convolutional Neural Network (CNN) is employed for pointwise classification of image pixels and the dependence of classification results on the type of input color channels and grapes color properties are studied. The advantage of using additional perception-based input features, such as luminance and visual contrast, is also evaluated, as well as the dependence of the method on the choice of the training set in terms of the amount of labeled data. The latter point has a significant impact on the practical use of the method on-site, its usability by non-expert users, and its adaptability to individual vineyards. Experimental results show that a properly trained CNN can discriminate and detect grape bunches even under uncontrolled acquisition conditions and with limited computational load, making the proposed method implementable on smart devices and suitable for on-site and real-time applications.

This Special Issue of Applied Numerical Mathematics contains selected and refereed papers presented at the 21st IMACS World Congress which was held during September 11–15, 2023 in Rome, Italy.

We present a new image scaling method both for downscaling and upscaling, running with any scale factor or desired size. The resized image is achieved by sampling a bivariate polynomial which globally interpolates the data at the new scale. The method's particularities lay in both the sampling model and the interpolation polynomial we use. Rather than classical uniform grids, we consider an unusual sampling system based on Chebyshev zeros of the first kind. Such optimal distribution of nodes permits to consider near-best interpolation polynomials defined by a filter of de la Vallée-Poussin type. The action ray of this filter provides an additional parameter that can be suitably regulated to improve the approximation. The method has been tested on a significant number of different image datasets. The results are evaluated in qualitative and quantitative terms and compared with other available competitive methods. The perceived quality of the resulting scaled images is such that important details are preserved, and the appearance of artifacts is low. Competitive quality measurement values, good visual quality, limited computational effort, and moderate memory demand make the method suitable for real-world applications.

Image scaling methods allows us to obtain a given image at a different, higher (upscaling) or lower (downscaling), resolution with the aim of preserving as much as possible the original content and the quality of the image. In this paper, we focus on interpolation methods for scaling three-dimensional grayscale images. Within a unified framework, we introduce two different scaling methods, respectively based on the Lagrange and filtered de la Vall\'ee Poussin type interpolation at the 1st kind's Chebyshev zeros. In both cases, using a non-standard sampling model, we take (via tensor product) the associated trivariate polynomial interpolating the input image. It represents a continuous approximate 3D image to resample at the desired resolution. Using discrete linf and l2 norms, we theoretically estimate the error achieved in output, showing how it depends on the error in input and on the smoothness of the specific image we are processing. Finally, taking the special case of medical images as a case study, we experimentally compare the performances of the proposed methods among them and with the classical multivariate cubic and Lanczos interpolation methods.

Image resizing (IR) has a crucial role in remote sensing (RS), since an image's level of detail depends on the spatial resolution of the acquisition sensor; its design limitations; and other factors such as (a) the weather conditions, (b) the lighting, and (c) the distance between the satellite platform and the ground targets. In this paper, we assessed some recent IR methods for RS applications (RSAs) by proposing a useful open framework to study, develop, and compare them. The proposed framework could manage any kind of color image and was instantiated as a Matlab package made freely available on Github. Here, we employed it to perform extensive experiments across multiple public RS image datasets and two new datasets included in the framework to evaluate, qualitatively and quantitatively, the performance of each method in terms of image quality and statistical measures.

Resolution Approximation Methods (RAM) play a crucial role in many real-world applications where preserving the original image quality is essential. Depending on the specific applicative field, the approximation may focus on spatial and/or color (intensity) information [7], [6]. Over the years, several methods have been proposed for color (gray) images, and multiple research directions have been pursued to enhance the performance and robustness of RAM [1],[2], [3], [4] and [5]. This contribution explores some approaches for both spatial and color (intensity) resolution approximation, providing a comprehensive analysis of their benefits, drawbacks, and potential future advancements.

Smart farming is becoming an active and interdisciplinary research field as it requires to solve interesting and challenging research issues to respond concretely to the demands of specific use-cases. One of the most delicate tasks is the automatic yield estimation, as for example in vineyards [1]. Computer vision methods that implement the rules of the human visual system can contribute to task accomplishment as they simulate what winemakers make manually [2]. An automatic artificial-intelligence method for grape bunch detection from RGB images is presented. It properly defines the input of a Convolutional Neural Network whose task is the segmentation of grape bunches [3]. The network input consists of pointwise visual contrast-based measurements that allow us to discriminate and detect grape bunches even in uncontrolled acquisition conditions and with limited computational load. The latter property makesthe proposed method implementable on smart devices and appropriate for onsite and real-time applications.

Image resizing is a basic tool in image processing, and in literature, we have many methods based on different approaches, which are often specialized in only upscaling or downscaling. In this paper, independently of the (reduced or enlarged) size we aim to get, we approach the problem at a continuous scale where the underlying function representing the image is globally approximated by its Lagrange-Chebyshev I kind interpolation polynomial corresponding to suitable (tensor product) grids of first kind Chebyshev zeros. This is a well-known approximation tool widely used in many applicative fields due to the optimal behavior of the related Lebesgue constants. Here we aim to show how Lagrange-Chebyshev interpolation can be fruitfully applied also for resizing any digital image in both downscaling and upscaling at any desired size. The performance of the proposed method has been tested in terms of the standard SSIM (Structured Similarity Index Measurement) and PSNR (Peak Signal to Noise Ratio) metrics. The results indicate that, in upscaling, it is almost comparable with the classical Bicubic resizing method with slightly better metrics, but in downscaling a much higher performance has been observed in comparison with Bicubic and other recent methods too. Moreover, for all downscaling cases with an odd scale factor, we give a theoretical estimate of the MSE (Mean Squared Error) of the output image produced by our method, stating that it is certainly null (hence PSNR equals infinite and SSIM equals one) if the input image's MSE is null.

Skin lesion segmentation is one of the crucial steps for an efficient non-invasive computer-aided early diagnosis of melanoma. This paper investigates how to use colour information, besides saliency, for determining the pigmented lesion region automatically. Unlike most existing segmentation methods using only the saliency to discriminate against the skin lesion from the surrounding regions, we propose a novel method employing a binarization process coupled with new perceptual criteria, inspired by the human visual perception, related to the properties of saliency and colour of the input image data distribution. As a means of refining the accuracy of the proposed method, the segmentation step is preceded by a pre-processing aimed at reducing the computation burden, removing artefacts, and improving contrast. We have assessed the method on two public databases, including 1497 dermoscopic images. We have also compared its performance with classical and recent saliency-based methods designed explicitly for dermoscopic images. The qualitative and quantitative evaluation indicates that the proposed method is promising since it produces an accurate skin lesion segmentation and performs satisfactorily compared to other existing saliency-based segmentation methods.

The aim of this talk is to show how de la Vallee Poussin type interpolation based on Chebyshev zeros of rst kind, can be applied to resize an arbitrary color digital image. In fact, using such kind of approximation, we get an image scaling method running for any desired scaling factor or size, in both downscaling and upscaling. The peculiarities and the performance of such method will be discussed.

In a computer-aided system for skin cancer diagnosis, hair removal is one of the main challenges to face before applying a process of automatic skin lesion segmentation and classification. In this paper, we propose a straightforward method to detect and remove hair from dermoscopic images. Preliminarily, the regions to consider as candidate hair regions and the border/corner components located on the image frame are automatically detected. Then, the hair regions are determined using information regarding the saliency, shape and image colors. Finally, the detected hair regions are restored by a simple inpainting method. The method is evaluated on a publicly available dataset, comprising 340 images in total, extracted from two commonly used public databases, and on an available specific dataset including 13 images already used by other authors for evaluation and comparison purposes. We propose also a method for qualitative and quantitative evaluation of a hair removal method. The results of the evaluation are promising as the detection of the hair regions is accurate, and the performance results are satisfactory in comparison to other existing hair removal methods.

The visual quality evaluation is one of the fundamental challenging problems in image processing. It plays a central role in the shaping, implementation, optimization, and testing of many methods. The existing image quality assessment methods centered mainly on images altered by common distortions while paying little attention to the distortion introduced by color quantization. This happens despite there is a wide range of applications requiring color quantization as a preprocessing step since many color-based tasks are more efficiently accomplished on an image with a reduced number of colors. To fill this gap, at least partially, we carry out a quantitative performance evaluation of nine currently widely-used full-reference image quality assessment measures. The evaluation runs on two publicly available and subjectively rated image quality databases for color quantization degradation by considering their appropriate combinations and subparts. The evaluation results indicate what are the quality measures that have closer performances in terms of their correlation to the subjective human rating and prove that the selected image database significantly impacts the evaluation of the quality measures, although a similar trend on each database is maintained. The detected strong trend similarity, both on individual databases and databases obtained by a proper combination, provides the ability to validate the database combination process and consider the quantitative performance evaluation on each database as an indicator for performance on the other databases. The experimental results are useful to address the choice of appropriate quality measures for color quantization and to improve their future employment.

Analisi, modellizzazione e simulazione delle traiettorie dei visitatori nelle aree museali

Segmenting skin lesions in dermoscopic images is a key step for the automatic diagnosis of melanoma. In this framework, this paper presents a new algorithm that after a pre-processing phase aimed at reducing the computation burden, removing artifacts and improving contrast, selects the skin lesion pixels in terms of their saliency and color. The method is tested on a publicly available dataset and is evaluated both qualitatively and quantitatively.

In this paper, an adaptive method for copy-move forgery detection and localization in digital images is proposed. The method employs wavelet transform with non constant Q factor and characterizes image pixels through the multiscale behavior of corresponding wavelet coefficients. The detection of forged regions is then performed by considering similar those pixels having the same multiscale behavior. The method is pointwise and the length of pixel features vector is image dependent, allowing for a more precise and fast detection of forged regions. The qualitative and quantitative evaluation of the experimental results reveals that the proposed method outperforms some existing transform-based methods in terms of performance and execution time.

The paper presents a method for color quantization (CQ) which uses visual contrast for determining an image-dependent color palette. The proposed method selects image regions in a hierarchical way, according to the visual importance of their colors with respect to the whole image. The method is automatic, image dependent and requires a moderate computational effort. Preliminary results show that the quality of quantized images, measured in terms of Mean Square Error, Color Loss and SSIM, is competitive with some existing CQ approaches.

A new technique for color quantization is suggested. First, pre-quantization is accomplished by means of spatial resolution reduction; then, color aggregation is accomplished based on the distance between colors in the color space. Color aggregation is an iterated process where the number of iterations is given by the difference between the number of colors of the pre-quantized image, and the number of colors desired for the quantized image. Color mapping is finally accomplished. Performance evaluation is done in terms of generally adopted quality measures. Comparisons with other methods in the literature are also provided.