Detecting Symbols in Images

Optičko prepoznavanje znakova (OCR) je grana računalnog vida koja se bavi raspoznavanjem znakova na slikama. Proces OCR sustava započinje predobradom slike tako da postane pogodna za jasnu segmentaciju znakova. Predobrada slike uključuje uklanjanje i zaglađivanje smetnji, otkrivanje i ispravak nagiba dokumenta, normalizaciju. Segmentiranim znakovima se izvlače njihove značajke te se na temelju njih znak provodi kroz raspoznavanje. Značajke mogu biti statističke i strukturne značajke te globalne transformacije. Popularna i učinkovita metoda raspoznavanja znakova je konvolucijska neuronska mreža.Optical character recognition is a branch of computer vision that is focused on recognizing characters in images. The OCR system process begins with image preprocessing to make the image suitable for clear character segmentation. Image preprocessing includes noise removal and smoothing, document skew detection and correction, normalization. Features are extracted from the segmented characters, and based on them the character is being recognized. Features can be statistical and structural features, and global transformations. A popular and effective method of character recognition is the convolutional neural network