Numerous efforts have been done for text location in videos and pictures. Candidate text regions are located by corner points detection. “Corner” is a 2D point in a picture has high curvature with the region boundary. deleting isolate corners that have few neighbor corners, and merging the remaining corners if they are close to each other. Candidate text regions are then decomposed to get candidate text lines by similar method to the projection profiles analysis that utilizes vertical and horizontal sobel edge maps of the candidate text regions. At last, eliminate some false alarms by making text box verification based on Fill Factor, and Center Offset Ratio Constraints that also utilizing sobel vertical and horizontal edge maps of the detected …show more content…
The horizontal and vertical edges of an image are computed. Edges are connected using dilation. The vertical edges are connected in the horizontal direction. The horizontal edges are connected in the vertical direction. Localization of text is refined by detecting text lines. Text lines are extracted from the text blocks by recognizing the baselines below and above the text. Text image is segmented. The gray scale image is calculated. For every pixel in the detected text boxes a vector composed of the entropy and standard deviation of its eight neighbors is calculated. Fuzzy C-means clustering algorithm is utilized to binarize the text picture [5].

2.1.1.3 Texture-Based Methods
Texture-based considers text as a special texture and thus, extract text by utilizing conventional texture classification methods and uses various techniques for feature extraction. these methods require extensive training as it is sensitive to font styles and sizes and are computationally costly for big databases[1][2][3] [6].
2.1.1.4 Correlation-Based Methods the correlation based techniques utilize any correlation method to decide if a pixel belongs to a text area or not