Medical image segmentation is a significant and complex challenge in medical imaging. In recent years, deep learning models have been applied to image segmentation and have shown exceptional performance. However, medical image segmentation has a scarcity of expert-labeled data compared to other deep learning research fields. Therefore, augmenting medical expert-labeled data are primarily the easiest and fastest way to improve the deep learning model’s performance. In this paper, computed tomography image segmentation technology based on the ResNet network integrated into the probabilistic model has been proposed. The proposed segmentation technology is based on the deep learning model of the ResNet50 architecture to extract features from images and initially detect objects of interest and on the probabilistic model with weighted parameters that employs conditional random fields, the GrabCut algorithm, and the argmax function to perform the final detection of objects of interest.
To train, test, and evaluate the effectiveness of the proposed method, appropriate chest CT datasets were identified to solve the task of segmenting the lung cavity, the liver and areas affected by COVID-19. The proposed image segmentation technology demonstrates segmentation accuracy results of 73.12% by Dice Score for the COVID-19 disease dataset, 97.71% for the lung cavity dataset, and 98.36% for the liver dataset, which perform better than state-of-the-art solutions.
The proposed image segmentation technology has been compared with state-of-the-art technologies (SegNet, UNet, and FCN-ResNet50) for CT segmentation to demonstrate the effectiveness of the method. The positive outcomes strongly suggest the significant potential of the proposed image segmentation technology. According to the obtained results, the proposed image segmentation technology can be a useful auxiliary tool for doctors to segment CT images for further analysis and monitoring of statistical and dynamic indicators.

Low-light scenes are characterized by the loss of illumination, the noise, the color distortion and serious information degradation. The low-light image enhancement is a significant part of computer vision technology. The low-light image enhancement methods aim to an image recover to a normal-light image from dark one, a noise-free image from a noisy one, a clear image from distorting one. In this paper, the low-light image enhancement technology based on Retinex-based deep network combined with the image processing-based module is proposed. The proposed technology combines the use of traditional and deep learning methodologies, designed within a simple yet efficient architectural framework that focuses on essential feature extraction. The proposed preprocessing module of low-light image enhancement is centered on the unique knowledge and features of an image. The choice of a color model and a technique of an image transformation depends on an image dynamic range to ensure high results in terms of transfer a color, detail integrity and overall visual quality. The proposed Retinex-based deep network has been trained and tested on transformed images by means of preprocessing module that leads to an effective supervised approach to low-light image enhancement and provide superior performance. The proposed preprocessing module is implemented as an independent image enhancement module in a computer system of an image analysis and as the component module in a neural network system of an image analysis. Experimental results on the low light paired dataset show that the proposed method can reduce noise and artifacts in low-light images, and can improve contrast and brightness, demonstrating its advantages. The proposed approach injects new ideas into low light image enhancement, providing practical applications in challenging low-light scenarios.

A colorization method based on a fully convolutional neural network for grayscale images is presented in this paper. The proposed colorization method includes color space conversion, grayscale image preprocessing and implementation of improved U-Net network. The training and operating of the U-Net network take place for images represented in the space of the Lab color model. The trained U-Net network integrates realistic colors (generate data of a and b components) into grayscale images based on L-component data of the Lab color model. Median cut method of quantization is applied to L-component data before the training and operating of the U-Net network. Logistic activation function is applied to normalized results of convolution layers of the U-Net network. The proposed colorization method has been tested on ImageNet database. The evaluation results of the proposed method according to various parameters are presented. Colorization accuracy by the proposed method reachers more than 84.81%. The colorization method proposed in this paper is characterized by optimized architecture of convolution neural network that is able to train on a limited image set with a satisfactory training duration. The proposed colorization method can be used to improve the image quality and restoring data in the development of computer vision systems. The further research can be focused on the study of a technique of defining optimal number of the gray levels and the implementation of the combined quantization methods. Also, further research can be focused on the use of HSV, HLS and other color models for the training and operating of the neural network.