International Journal of Image and Graphics

Mammography imaging is one of the most successful techniques for breast cancer screening and detecting breast lesions. Detection of the Region of Interest (ROI) (where the possible abnormalities could be present) is the backbone for the success of any Computer-Aided Detection or Diagnosis (CADx) system. In this paper, to assist the CADx system, one computational model is proposed to detect breast mass lesions from mammogram images. At the beginning of the process, pectoral muscles from the mammograms are removed as a pre-processing step. Then by applying an automatic thresholding scheme with the required image processing techniques, different regions of breast tissues are ranked to detect the possible suspected region to refine the further segmentation task. One seeded region growing approach is proposed with an automatic seed selection criterion to detect the suspected region to segment the ROI. The proposed model has very less user intervention as maximum of the parameters are computed automatically. To evaluate the performance of the proposed model, it is compared with four different methods with six different evaluation metrics viz., Jaccard & Dice co-efficient, relative error, segmentation accuracy, error and Fowlkes–Mallows index (FMI). On the proposed model, 57 mammogram images are tested, consisting of four different cases that are collected from the publicly available benchmark database. The qualitative and quantitative analyses are performed to evaluate the proposed model. The best dice co-efficient, Jaccard co-efficient, accuracy, error and FMI values observed are 0.9506, 0.9471, 95.62%, 4.38% and 0.932, respectively. The superiority of the model over six state-of-the-art compared methods is well evident from the experimental results.

The prominence and popularity of Image Denoising in medical image processing has been obvious since its early conception. Medical Image Denoising is primarily a significant pre-processing method for further image processing steps in various fields. Its ability to speed up the diagnosis by enhancing the sensory quality of noisy images is proven to be working in most of the cases. The efficiency of the deep neural networks for Medical Image Denoising has been well proven traditionally. Both noisy and clean images are equal requirements in most of these training methods. However, it is not always possible to procure clean images for various applications such as Dynamic Imaging, Computed Tomography, Magnetic Resonance Imaging, and Camera Photography due to the inevitable presence of naturally occurring noisy signals which are intrinsic to the images. There have been self-supervised single Image Denoising methods proposed recently. Being inspired by these methods, taking this a step further, we propose a novel and better denoising method for single images by training the learning model on each of the channels of the input data, which is termed as “Noise2Split”. It ultimately proves to reduce the noise granularly in each channel, pixel by pixel, by using Single Channeled Patch-Based (SCPB) learning, which is found to be resulting in a better performance. Further, to obtain optimum results, the method leverages BRISQUE image quality assessment. The model is demonstrated on X-ray, CT, PET, Microscopy, and real-world noisy images.

As innovations for image handling, image enrichment (IE) can give more effective information and image compression can decrease memory space. IE plays a vital role in the medical field for which we have to use a noiseless image. IE applies to all areas of understanding and analysis of images. This paper provides an innovative algorithm called contrast-limited adaptive fuzzy gamma (CLAFG) for IE using chest X-ray (CXR) images. The image dissimilarity is enriched by computing several histograms and membership planes. The proposed algorithm comprises various steps. Firstly, CXR is separated into contextual region (CR). Secondly, the cliplimit, a threshold value which alters the dissimilarity of the CXR and applies it to the histogram which, is generated by CR and then applies the fuzzification technique via the membership plane to the CXR. Thirdly, the clipped histograms are performed in two ways, i.e. it is merged using bi-cubic interpolation techniques and it is modified with membership function. Finally, the resulting output from bi-cubic interpolation and membership function are fond of using upgrade contemplate standard methods for a richer CXR image.

The life of the human beings are considered as the most precious and the average life time has reduced from 75 to 50 age over the past two decades. This reduction of average life time is due to various health hazards namely cancer and many more. The brain tumor ranks among the top ten most common source of demise. Although brain tumors are not the leading cause of death globally, 40% of other cancers (such as breast or lung cancers) metastasize to the brain and become brain tumors. Despite being the gold norm for tumor diagnosis, a biopsy has a number of drawbacks, including inferior sensitivity/specificity, and menace when performing the biopsy, and lengthy wait times for the results. This work employs artificial intelligence integrated with the honey bee optimization (HBO) in detecting the brain tumor with high level of execution in terms of accuracy, recall, precision, F1 score and Jaccard index when compared to the deep learning algorithms of long short term memory networks (LSTM), convolutional neural networks, generative adversarial networks, recurrent neural networks, and deep belief networks. In this work, to enhance the level of prediction, the image segmentation methodology is performed by the probabilistic U-RSNet. This work is analyzed employing the BraTS 2020, BraTS 2021, and OASIS dataset for the vital parameters like accuracy, precision, recall, F1 score, Jaccard index and PPV.

Vision transformers are deep neural networks applied to image classification based on a self-attention mechanism and can process data in parallel. Aiming at the structural loss of Vision transformers, this paper combines ConViT and Convolutional Neural Network (CNN) and proposes a new model Convolution Meet Vision Transformers (CMVT). This model adds a convolution module to the ConViT network to solve the structural loss of the transformer. By adding hierarchical data representation, the ability to gradually extract more image classification features is improved. We have conducted comparative experiments on multiple dataset, and all of them have been enhanced to improve the efficiency and performance of the model.

In remote sensing, owing to existing sensors’ limitations and the tradeoff between signal-to-noise ratio (SNR) and instantaneous field of view (IFOV), it is difficult to obtain a single image with good spectral and spatial resolution. Pansharpening (PS) is the technique for sharpening multispectral (MS) images by extracting structural and edge information of panchromatic (PAN) image. Multiscale decomposition methods are used for decomposing image in sub-bands but are affected by ringing artifacts, therefore the resultant image seems to be blurred and misregistered. The proposed method overcomes this drawback by decomposing PAN and four band MS image into cartoon and texture components with total variation (TV) Hilbert${}^{-1}$ model. The particle swarm optimization (PSO) algorithm is used for finding the optimum weight for fusing texture and cartoon details of PAN and MS images. The proposed method is practically validated on both full-scale and reduced-scale. Robustness of our proposed approach is tested on different geographical areas such as hilly, urban, and vegetation areas. From the visual analysis and qualitative parameters, the proposed method is proved effective compared with other traditional approaches.

Recent years have seen a significant amount of studies in the area of speech enhancement. This review looks at several speech improvement methods as well as Deep Neural Network (DNN) functions in speech enhancement. Speech transmissions are frequently distorted by ambient noise, background noise, and reverberations. There are processing methods, such as Short-time Fourier Transform, Short-time Autocorrelation, and Short-time Energy (STE), that can be used to enhance speech. To reduce speech noise, features such as the Mel-Frequency Cepstral Coefficients (MFCCs), Logarithmic Power Spectrum (LPS), and Gammatone Frequency Cepstral Coefficients (GFCCs) can be retrieved and input to a DNN. DNN is essential to speech improvement since it builds models using a lot of training data and evaluates the efficacy of the enhanced speech using certain performance metrics. Since the beginning of deep learning publications in 1993, a variety of speech enhancement methods have been examined in this study. This review provides a thorough examination of the several neural network topologies, training algorithms, activation functions, training targets, acoustic features, and databases that were employed for the job of speech enhancement and were gathered from various articles published between 1993 and 2022.

Image editing technologies have been advanced that can significantly enhance the image, but can also be used maliciously. Colorization is a new image editing technology that uses realistic colors to colorize grayscale photos. However, this strategy can be used on natural color images for a malicious purpose (e.g. to confuse object recognition systems that depend on the colors of objects for recognition). Image forensics is a well-developed field that examines photos of specified conditions to build confidence and authenticity. This work proposes a new fake colorized image detection approach based on the special Residual Network (ResNet) architecture. ResNets are a kind of Convolutional Neural Networks (CNNs) architecture that has been widely adopted and applied for various tasks. At first, the input image is reconstructed via a special image representation that combines color information from three separate color spaces (HSV, Lab, and Ycbcr); then, the new reconstructed images have been used for training the proposed ResNet model. Experimental results have demonstrated that our proposed method is highly generalized and significantly robust for revealing fake colorized images generated by various colorization methods.

In this paper, we present a denoising method for path tracing using residual learning with convolutional neural networks (CNNs). Noisy artifacts in path tracing are inherited from insufficient sampling, which often generates over- or under-exposed values when integrating the limited bright or dark samples in a pixel. In this paper, we introduce a dual channel residual learning CNNs which separates the over and under-exposed signals in order to provide an efficient denoising filter for the path tracing rendering. Furthermore, we present an advanced CNN comprised of variable-sized kernels in each convolutional layer. Our CNN detects features in different scales providing an adaptive denoising filter capability which is optimal for extracting various contextual details in a complex scene. The experiments demonstrate that our method generates better visual quality than other compared approaches across various rendering effects.

Medical image classification is one of the most significant tasks in computer-aided diagnosis. In the era of modern healthcare, the progress of digitalized medical images has led to a crucial role in analyzing medical image analysis. Recently, accurate disease recognition from medical Computed Tomography (CT) images remains a challenging scenario which is important in rendering effective treatment to patients. The infectious COVID-19 disease is highly contagious and leads to a rapid increase in infected individuals. Some drawbacks noticed with RT-PCR kits are high false negative rate (FNR) and a shortage in the number of test kits. Hence, a Chest CT scan is introduced instead of RT-PCR which plays an important role in diagnosing and screening COVID-19 infections. However, manual examination of CT scans performed by radiologists can be time-consuming, and a manual review of each individual CT image may not be feasible in emergencies. Therefore, there is a need to perform automated COVID-19 detection with the advances in AI-based models. This work presents effective and automatic Deep Learning (DL)-based COVID-19 detection using Chest CT images. Initially, the data is gathered and pre-processed through Spatial Weighted Bilateral Filter (SWBF) to eradicate unwanted distortions. The extraction of deep features is processed using Fine_Dense Convolutional Network (Fine_DenseNet). For classification, the Softmax layer of Fine_DenseNet is replaced using Improved Generative Adversarial Network_Artificial Hummingbird (IGAN_AHb) model in order to train the data on the labeled and unlabeled dataset. The loss in the network model is optimized using Artificial Hummingbird (AHb) optimization algorithm. Here, the proposed DL model (Fine_DenseIGANet) is used to perform automated multi-class classification of COVID-19 using CT scan images and attained a superior classification accuracy of 95.73% over other DL models.