Academia.eduAcademia.edu

Figure 3 – uploaded by Abdolhossein Sarrafzadeh

See full PDFdownloadDownload figure
Fig. 3. (a) A document image with 5 degree skew, (b) result of performing closing on the (a) with 70x30 structural element. In this method, the image text lines are extracted as suitable features for skew detection using morphological operators. In this method, fill the space between the characters by using morphological operators. Thus, any text lines are black blocks. One of the morphological operators which can be used for this purpose is closing. For example, a document image with 5 degrees skew and a result of performing closing with a 70x30 structural element are shown in Fig. 3.

Figure 3 (a) A document image with 5 degree skew, (b) result of performing closing on the (a) with 70x30 structural element. In this method, the image text lines are extracted as suitable features for skew detection using morphological operators. In this method, fill the space between the characters by using morphological operators. Thus, any text lines are black blocks. One of the morphological operators which can be used for this purpose is closing. For example, a document image with 5 degrees skew and a result of performing closing with a 70x30 structural element are shown in Fig. 3.

Related Figures (2)

Whenever the value of f(x,yo) and f(x +s,yo+d) grow larger, the result of their multiplication is greater. Whenever the multiplication of f(x, yo) and f(x + 5,¥9 + d) for all values of x becomes larger, the value of cross- correlation function between the two lines is greater. Large values of f(x, ¥9) and f(x +s,y¥o +d) correspond to the black triangle in distance between two text lines, and the other side is d the he two imaginary vertical lines. Angle 0 shown in the triangle is the angle of the text line with the distance between Taking the ight image pixels. Most of the light image pixels are the pixels between the lines. So, if for most values of x, f(x, yo) and f(x + 5s,Y 9 + d) correspond to the light pixe he lines, the cross-correlation function between the two lines is greater. This happens when s is equal to the distance between two text lines. It is clear in Fig. 1. distance between two text lines and between two vertical lines, the text line angle from the horizontal axis can be found. To understand how to calculate this ang s between imaginary e, note the Fig. 2. One side of the triang e is s the horizontal axis and can be calculated as arctan(s/d). This angle is the skew of document image. Thus, the cross- correlation function can he used to detect the skew.
Whenever the value of f(x,yo) and f(x +s,yo+d) grow larger, the result of their multiplication is greater. Whenever the multiplication of f(x, yo) and f(x + 5,¥9 + d) for all values of x becomes larger, the value of cross- correlation function between the two lines is greater. Large values of f(x, ¥9) and f(x +s,y¥o +d) correspond to the black triangle in distance between two text lines, and the other side is d the he two imaginary vertical lines. Angle 0 shown in the triangle is the angle of the text line with the distance between Taking the ight image pixels. Most of the light image pixels are the pixels between the lines. So, if for most values of x, f(x, yo) and f(x + 5s,Y 9 + d) correspond to the light pixe he lines, the cross-correlation function between the two lines is greater. This happens when s is equal to the distance between two text lines. It is clear in Fig. 1. distance between two text lines and between two vertical lines, the text line angle from the horizontal axis can be found. To understand how to calculate this ang s between imaginary e, note the Fig. 2. One side of the triang e is s the horizontal axis and can be calculated as arctan(s/d). This angle is the skew of document image. Thus, the cross- correlation function can he used to detect the skew.
In 1993, Yan used cross-correlation function skew detection for the first time [22]. Yan’s method is as follows:
In 1993, Yan used cross-correlation function skew detection for the first time [22]. Yan’s method is as follows:
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved