Microsoft Word Avtoreferat (Muhiddinov M. N. Tatu)

Connection of dissertation topic with the plans of research activities of 
the higher education institution where the dissertation is carried out. 
The 
research was carried out within the framework of the projects on number №BV-
F4-011 «Development of theory, methods and tools of parallel computing in signal 
processing and image processing» (2017-2019), №BV-Atex-2018-251 
«Development of Uzbek language Text-To-Speech (TTS) software and speech 
synthesizer for the visually impaired to use computer technology to read and to 

---

23 
write texts» (2018-2019) of the Tashkent University of Information Technologies 
named after Muhammad al-Khwarizmi. 
The aim of the research work
is a development of methods, algorithms and 
software of salient objects extraction, its edge detection and scene text recognition, 
pronunciation from natural scene images. 
The tasks of the research work:
analysis of methods of salient objects extraction in images based on global 
contrast enhancement of the color values with locally adaptive triple thresholding; 
development of algorithmic method of Uzbek language texts detection, 
extraction and recognition in natural images based on fully convolutional neural 
network model; 
development of algorithm and software of graphic objects extraction and 
inner and outer edge detection in images to represent on tactile graphics; 
development of algorithm and software of Uzbek language texts 
pronunciation in natural images for perception process of visually impaired. 
The object of the research
work
are natural scene images, video 
information, words and audio information for Uzbek language text-to-speech 
synthesis as well as image processing tools. 
The subject of the research work
– digital image processing algorithms, 
neural network models, and programming libraries to solve the problem of object 
detection effectively, extraction and text recognition in images.