Abstract:

Abstract Context. Presence of fog and haze on digital images may cause problems in processes of recognition, tracking, classification of objects. Thus methods for removing fog and improving visibility of objects in images obtained under poor visibility conditions are in demand in many computer vision problems. In foggy weather, contrast and color of an image get worse. Fog removal is often accompanied by artifacts in the image and color distortion. Therefore, it is relevant to seek methods for correct assessing presence and removal of fog while preserving image details and colors and developing appropriate methods for blurred images processing. Objective. The purpose of this research is to find effective approaches to solving the problem of removing fog and haze from digital images and implementing them in a digital image processing computer system [1]. Method. Main stages of image processing are performed on the intensity channel, which helps to preserve colors. The proposed approach keeps the values of the processed pixels in an acceptable range, which allows better preservation of image details. Frequency filters are used to evaluate the transmission map. In a modified method, fog density is estimated using a neural network. Results. The method of removing fog and haze from single image is proposed. This method effectively improves the objects visibility, preserves details and colors in the image. A modification of the method with another fog density estimation method is also proposed. The presented methods were implemented in a computer system [1]. Conclusions. The proposed method and its modification effectively remove fog and haze from single image and improve the objects distinguishability in them. The implementation of these methods in a computer image processing system [1] has expanded the functionality of the system and increased its ability to improve the quality of images obtained under poor visibility conditions. The system can be used for preliminary image processing to prevent errors in further operation of computer vision algorithms. Author Biographies М. Е. Serdiuk, Oles Honchar Dnipro National University, Dnipro PhD, Associate Professor of Computer Technologies Department V. G. Berkut, Oles Honchar Dnipro National University, Dnipro Master of Applied Mathematics, graduate student of Applied Mathematics Faculty S. F. Sirik, Oles Honchar Dnipro National University, Dnipro Assistant of Computer Technologies Department References Serdiuk М. Е., Berkut V. G. Kompiuterna systema lokalizatsii ta vydalennia tinei u tsyfrovykh zobrazhenniakh, Radio Electronics,

Keywords: