aAdvanced interpretation mode algorithm
INTERPRETATION MODE ALGORITHM
In our application, a blind user first takes a picture, then our system automatically detects text areas in the picture and delivers the layout of the document. Finally, text can be transformed into speech signal and their major constraints are the Text image deterioration and low computational resources. The first is detection and localization of the text regions. The idea is to locate the text elements without necessarily recognizing them, cut them out of the image, determine the reading order and finally correct their perspective. The read mode uses the principles of OCR, which deals with the recognition of the printed text and storing it in the coded standards the most intuitive characteristics of text are its regularity. Printed text consists of characters with approximately the same size and line thickness that are located at a regular distance from each other.
Such regularities can also be observed from edges being detected on textual boundaries Text shows spatial cohesionâ€”characters of the same text string (a word, or words in the same line) are of similar heights, orientation, and spacing. Characters contrast with their background since they are designed to be read easily. Characters appear in clusters at a limited distance aligned to a virtual line. Most of the time the orientation of these virtual lines is horizontal since that is the conventional way of writing. Most of the previous researches about text detection focus on extracting text from video. Techniques applied to images or video key frames can broadly be classified as edge, color or texture based.Each approach has its advantages /drawbacks concerning accuracy, efficiency and computational requirements. Edge based techniques use edges information in order to characterize text areas. Edges of text symbols are typically stronger than those of noise or background areas. These methods operate essentially in gray scale format and do not require much processing time.Nevertheless, they do not cope with complex text images like pictures of magazines or scene images where edge information alone is not sufficient to separate text from a noisy background. The use of color information allows the image to be segmented into connected components of uniform color. A reduction of the color palette is often required.