Sie sind auf Seite 1von 29

Hybrid Halftoning, A Useful Method for Flexography

JOURNAL OF IMAGING SCIENCE AND TECHNOLOGY vol. 49, no. 1, January/February 2005
Sasan Gooran

School of Electrical Engineering and Computer Science Kyungpook National Univ.

Abstract
Hybrid halftoning
Existing problem in the flexography using just an AM halftoning method
Not producing dots sufficiently small to handle the highlights and the shadows of the original image

Overcoming this problem


FM method in the highlights (and the shadows) AM method in the rest of the image

2 / 29

Introduction
Halftoning
AM halftoning
Being variable in the size of the dots while their spacing is constant

FM halftoning
Being kept constant in the size of the dots while their spacing varies

Replacement of the terminology


Not using the term halftone cell but using the term dpi

Flexography
Modified form of letterpress printing method
Usage of the packaging industry on the most varied materials
3 / 29

Lower print quality than that in offset printing Printing method that can print on very thin, flexible, and solid films, thick card boards, rough-surface packaging materials and fabrics

Problem
Not producing the dots sufficiently small
Handling the very light and dark parts of the original image by just using an AM method

Solution
Using AM halftoning in the mid-tones
Being lower dot gain than that of the FM halftoning Performing very well in the highlights by FM halftoning
Optimized dot positioning and the choice of the minimum dot size

4 / 29

Previous work
Present hybrid method
Not giving any detailed description on how the methods should be implemented Lin and Allebach
Using a stochastic dispersed-dot texture in highlights Utilizing a periodic clustered-dot pattern in mid-tones Not using the flexography but being electrographic devices Reducing the noisy impression in mid-tones

5 / 29

FM method
Process
Initial halftoned image
No dots

Solution of the problem of halftoning a grayscale image


Placing a number of dots on this empty initial image

Impression of lightness/darkness of the image


Being determined in advance for the number of dots to be placed
Sum of the pixel values in the original image rounded to the nearest integer

6 / 29

Where to place 1s (black dots)


Being placed for the dots iteratively
Decreasing the difference between the original image and the halftoned image

Being placed at the position of the darkest pixel in the original image for the first dot
Position of the maximum in the original image

Being low-pass filtered for the original image at first Feed-back process
Being placed for the first dot Being subtracted from the filter version of the original image for the filtered version of the current halftoned image Being founded for the position of the maximum Placing the second dot there Performing the feed-back process
7 / 29

Continuing until the predetermined number of dots are placing and the final image is achived

Not being reproduced for gray tones in the highlight and shadow regions
Overcoming the problem
Controlling the number of dots to be placed in a number of gray tone regions Using more control regions in these area Choosing seven control regions between 0 and 0.1 Choosing eight control regions between 0.1 and 0.9 Choosing seven control regions between 0.9 and 1 Totally 22 control regions Being determined for the number of dots in advance Sum of pixel values in the corresponding region

8 / 29

Problem in the extreme highlights


Not being placed homogeneously for the dots Cause
Size of filter that is chosen dependent on the gray tone The lighter the gray tone, the bigger the filter

Fig. 1.A constant image with a gray value of 1% is halftoned by the proposed FM method. a) The filter size is 11 11. b) The filter size is 21 21.
9 / 29

Hybrid halftoning
ppi, lpi and dpi
ppi (pixels per inch)
Scanning resolution
The number of samples taken per inch

lpi (lines per inch)


Line screen frequency
The number of halftone cells per inch Halftone cell Being divided into small areas for the original image

dpi (dots per inch)


Print resolution
The number of micro dots per inch Micro dots which consist of halftone cell
10 / 29

Size of the halftone cells


dpi lpi

The number of gray levels

(dpi

lpi ) + 1
2

(1)

Fig. 2. Two halftone cells are shown. The halftone cell to the left represents a gray value of 12 64 = 18.75% and the one to the right 16 64 = 25% .
11 / 29

The Hybrid halftoning method


Using Adobe Photoshop in order to AM halftone our images Representing black and white
1 and 0 respectively

Method
Being explained by both words and Matlab pseudo codes

Being known in advance


Critical dot ( its size and shape)
The smallest possible reproducible dot

Gray value for which AM halftoning method produce this dot


12 / 29

Independent components
Shape of halftone dots and halftone angle

Testing environment
Original test image g
600 600 pixels with ppi = 200, 600 dpi, 100 lpi , dot shape round,
o halftone angel 45, critical dot size 2 2 micro dots,

gray tone that corresponds to this dot: 0.125

Independent components
Halftone angle and the critical dot shape

Fig. 3. The original test image. 6 6 Pixels with ppi = 200 and printed at 200 dpi.
13 / 29

AM halftoning implemented process


Using Adobe Photoshop in order to AM halftone
g AM = AM _ halftone( g )

(2)

Building a mask using Eq.(3) in order to locate the highlights


mask = g < 0.125

where ppi = 2lpi and halftone cell is 6 6, g AM is 1800 1800 pixels

(3)

Meaning that mask is equal to 1 in positions where g is smaller than 0.125 and 0 elsewhere

Being AM halftoned exactly the same values as were used for AM halftoning the original image for mask
MASK = AM _ halftone(mask )

(4)

14 / 29

Fig. 4. The mask printed at 600 dpi. The parts that are black, i.e., 1, are the parts of the image that are going to be FM halftoned.The parts that are white, i.e.,0, are being AM-halftoned.

Using mask we can locate the parts of g AM where the dots are bigger than the critical size (bigger than 2 2) (5) g AM _ mask = ~ MASK & g AM
where ~ means a logical NOT and ~MASK is a matrix whose elements are 1s where MASK has zero elements, and 0s where MASK has nonzero elements. & denotes the element-by-element logical AND between the matrices

15 / 29

Fig. 5. An enlarged version of a part of g am _ mask is shown. There are some dots in the transition are that have been cut and are smaller than 2 2. Apart from these unwanted dots, the image also shows an enlarged version of a part of the test image that would be obtained in print if the print press were not able to print the dots smaller than 2 2. This image is printed at 150 dpi.

FM halftoning implemented method


Only using the critical dot, a 2 2 square dot Simplest way
Being first FM halftone the parts that should be FM halftoned, and then double the size of the result

16 / 29

Being resized to 900 900 pixels before being FM halftoned


g new = imresize( g ,1.5)

(6)

Only problem being MASK 1800 1800 and g new 900 900 (7) mask new = imresize(MASK , 0.5) Areas of the original image that are black in the mask that are supposed to be FM halftoned (8) g fm = FM _ halftone(g new & mask new ) Resizing again
G fm = imresize(g fm , 2 )

(9)

Removing the unwanted dots at boder between AM and FM halftoned parts in Gam Adding AM and FM parts together to get final hybrid halftoned image (10) Ghyb = G fm Gam
17 / 29

Fig. 6. Our test image and a grayscale ramp are halftoned by the proposed hybrid halftoning method. In the test image the parts that were black in MASK, shown in Fig. 4, are FM halftoned and the rest of the image AM halftoned. The FM method is non modified Error Diffusion. Image are printed at 600 dpi.

Fig. 7. Our test image and a grayscale ramp are halftoned by the proposed hybrid halftoning method. In the test image the parts that were black in MASK, shown in Fig. 4, are FM halftoned and the rest of the image AM halftoned. The FM method is the method presented in this article. Images and printed at 600 dpi.
18 / 29

How to remove the unwanted dots


Unwanted dots
Separating the AM part from the part being FM halftoned
Generating some dots in the transition
Being cut and consequently smaller than the critical dot

(a)

(b)

Fig. 8. The left half of the images is AM-halftoned and the right half is FM-halftoned. (a) the unwanted dots are not removed, hence a visible vertical dashed-line structure is observed in the border between AM and FM pattern. (b) The unwanted dots are removed and the visible vertical dashed-line has disappeared.

19 / 29

Reason for removing the unwanted dots


Degrading the quality of the final image

(a)

(b)

(c)

(d)

(e)

(f)

(g)

Fig. 9. How to remove the dots smaller than 2 2. The two dots to the right in a are going to be removed. The other two dots will remain as they are. (a) The original, called a. (b) The image b in Eq. (11). (c) The image c in Eq. (12). (d) The image d in Eq. (13). (e) The image e in Eq. (14). (f) The image f in Eq. (16). (g) The final image anew in Eq.(17).
20 / 29

Method
Using the convolution function in the Matlab Process
Locating the dots that are equal to the critical dot
Convoluting a with kernel that represents the critical dot 1 1 b = conv 2 a, 1 1
where conv 2 denotes the two-dimensional convolution and the matrix represents the 2 2 critical dot the kernel is not rotated if it has a symmetrical shape

(11)

21 / 29

Holding 4
The number of black micro dots in the critical dot Indication of a dot equal or bigger than the critical dot size (12) c = (b == 4 )
Meaning that matrix c is equal to 1 in those positions where b equals 4, and equal to 0 elsewhere

Convoluting c with the kernel


Not rotated version of the kernel

1 1 c, d = conv 2 , ' same' 1 1 where same makes the result d to be the same size as c

(13)

22 / 29

Rebuilding dots
Not some dots in the d but some dot in the original image

e = (a & ~ d ) 1 + (d ) 20 + (~ a & ~ d ) ( 1)

(14)

Finding connected micro black dot in d for e


1 1 1 , ' same' m = conv 2 e, 1 1 1 1 1 1

(15)

Black micro dots or connected to at least one black micro dots


f = m > 10 & e == 1

(16)

Final image
anew = f d

(17)
23 / 29

Critical Dots with General Shape


Process
Making a critical dot shape that is not square
Assuming to want the critical dot to be the dot in the upper left of Fig. 9(a) Being a 2 2 dot plus another three black micro dots

Making image p
First assuming the dot shape is 2 2 Assuming that (g new & mask new ) is constant image with 1% coverage Consisting of 7 black micro dots FM halftoning (4 7 g new & mask new ) Resizing it in order to 2 2

24 / 29

Detailed method
Finding the 2 2 dots
Replacing a by p in Eq.(11) and Eq.(12) Obtaining c

Computing the new dot


1 1 1 0 , ' same' q = conv 2 c, 0 1 1 1 0 0 1 0

(18)

Fig. 10. (a) The dots are 2 2, and (b) the dots are the new critical dot. The fractional are covered with black dots is the same as in a.

(a)

(b)

25 / 29

Test print
Process
Testing environment
Being ppi = 300, printing resolution: 2100 dpi , screen frequency: 150 lpi, gray levels=197, halftone angle= 37 o , and critical dot = 4 4

Comparison
Presented method and SambaFlex Result
Looking better in the highlights because the FM dots are placed more homogeneously

26 / 29

Fig. 11. An enlarged version of a part of the Fig. 12. An enlarged version of a part of the test print, printed at 300dpi. SambaFlex test print, printed at 300dpi. The hybrid raster method was used. method proposed in this article was used.

(a)

(b)

Fig. 13. (a) An enlarged version of a part of the image in Fig.11, printed at 150 dpi. (b) An enlarged version of a part of the image in Fig. 12, printed at 150 dpi.
27 / 29

Discussions and future work


Quality of halftoning method
Depending on the original images for real printing
Studying the mask before preparing the halftoned image

Result
Not suffering from any evident distortion in the transition area
Being used good FM method, the proposed method works very well

28 / 29

Very simple method


Needing only 17 lines for using Matlab
Well performing FM method

Future work
Extension of this hybrid halftonig method method to color
Halftoning the color separations dependently for the color extension of the FM method

29 / 29

Das könnte Ihnen auch gefallen