Microsoft Word Avtoreferat (Muhiddinov M. N. Tatu)

The methods of the research work.
In order to solve this problem modern 
methods of digital image processing, pattern recognition, character recognition, 
deep learning, probability theory, mathematical analysis, algorithmization, 
structured programming technology and performing computational experiments are 
used. 
The scientific novelty 
of the research work as follows: 
method of objects extraction in images based on global contrast enhancement 
of the color values with locally adaptive triple thresholding is improved; 
algorithmic method of Uzbek language texts detection, extraction and 
recognition in natural images based on fully convolutional neural network model is 
developed; 
algorithm and software of graphic objects extraction, inner and outer edge 
detection in images to represent on tactile graphics are developed; 
algorithm and software of Uzbek language texts pronunciation in natural 
images for perception process of visually impaired are developed. 
The practical results of the research work
are as follows: 
the software on detecting and extracting salient objects in natural scene 
images to generate tactile graphics for the visually impaired is developed; 
the software on detecting and recognizing text in natural scene images to 
generate Uzbek language TTS synthesizer for the visually impaired is developed; 
the proposed salient objects detection and extraction method have been 
implemented in the specialized boarding school for visually impaired № 77, Center 

---

24 
for the development of ICT in public education and used in tactile graphics 
generation process to give visual information for the visually impaired; 
the proposed scene text detection and recognition method have been 
implemented in the Central Board of the Uzbek Association of the Blind, the 
Center of youth and children with disabilities and applied in Uzbek language TTS 
synthesizer creation process for the visually impaired;