U.S. patent application number 09/812680 was filed with the patent office on 2001-08-23 for measuring image characteristics of output from a digital printer.
This patent application is currently assigned to I-Data International, Inc... Invention is credited to Banker, Yigal J., Monks, David E., Phillips, Lee E., Ting, David M.T..
Application Number | 20010016054 09/812680 |
Document ID | / |
Family ID | 21889295 |
Filed Date | 2001-08-23 |
United States Patent
Application |
20010016054 |
Kind Code |
A1 |
Banker, Yigal J. ; et
al. |
August 23, 2001 |
Measuring image characteristics of output from a digital
printer
Abstract
Measuring the printed image characteristics of printed output
from a digital printer by sending test pattern data to a digital
printer and generating a printed image of the test pattern data.
The printed image is scanned, and digital pattern data is output.
The digital pattern data is analyzed to generate one or more
quantitative ratings with respect to one or more printed image
characteristics.
Inventors: |
Banker, Yigal J.; (Chestnut
Hill, MA) ; Monks, David E.; (Lexington, MA) ;
Phillips, Lee E.; (Wellesley, MA) ; Ting, David
M.T.; (Sudbury, MA) |
Correspondence
Address: |
JACOBSON HOLMAN PLLC
400 SEVENTH STREET N.W.
SUITE 600
WASHINGTON
DC
20004
US
|
Assignee: |
I-Data International, Inc..
|
Family ID: |
21889295 |
Appl. No.: |
09/812680 |
Filed: |
March 21, 2001 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
09812680 |
Mar 21, 2001 |
|
|
|
09036563 |
Mar 9, 1998 |
|
|
|
Current U.S.
Class: |
382/112 |
Current CPC
Class: |
G06K 15/00 20130101;
H04N 1/00015 20130101; G06T 2207/10008 20130101; G06T 2207/30144
20130101; H04N 1/00076 20130101; H04N 1/00031 20130101; G06K
2215/0085 20130101; H04N 1/00053 20130101; G06T 7/001 20130101;
H04N 1/00058 20130101; H04N 1/00045 20130101; H04N 1/00002
20130101 |
Class at
Publication: |
382/112 |
International
Class: |
G06K 009/00 |
Claims
What is claimed is:
1. A method of measuring image characteristics of printed output
from a digital printer, comprising: sending test pattern data to a
digital printer, said test pattern including one or more target
objects designed to reveal one or more specific printed image
characteristics; generating a printed image of said test pattern
data at said digital printer; scanning said printed image to obtain
digital pattern data; and analyzing said digital pattern data to
generate one or more quantitative ratings with respect to said one
or more printed image characteristics.
2. The method of claim 1, wherein a printed test pattern is scanned
prior to sending said test pattern data to said digital
printer.
3. The method of claim 1, wherein said analyzing includes
generating a quality rating related to image quality from said one
or more quantitative ratings.
4. The method of claim 3, wherein said quality rating is based on a
human perception of quality.
5. The method of claim 1, further comprising: indicating
unacceptable print quality in response to said one or more
quantitative ratings.
6. The method of claim 1, further comprising: indicating a setting
adjustment to said digital printer in response to said one or more
quantitative ratings.
7. The method of claim 1, further comprising: automatically
adjusting a setting to said digital printer in response to said one
or more quantitative ratings.
8. The method of claim 1, wherein said analyzing step evaluates
said one or more quantitative ratings with respect to historical
data.
9. The method of claim 1, wherein said quantitative ratings include
a physical measurement of said one or more objects in said digital
pattern data.
10. The method of claim 3, wherein said scanner has higher
resolution than said digital printer.
11. The method of claim 9, wherein landmarks in said digital
pattern data are used to identify positions for said one or more
target objects.
12. The method of claim 11, wherein said landmarks are located in
the corners of said digital pattern data.
13. The method of claim 11, wherein said landmarks are used to
adjust for scanner deviations.
14. The method of claim 9, wherein said physical measurement
determines density uniformity.
15. The method of claim 9, wherein said physical measurement
determines positional accuracy.
16. The method of claim 9, wherein said physical measurement
determines edge sharpness.
17. The method of claim 9, wherein said physical measurement
determines edge acuity.
18. The method of claim 9, wherein said physical measurement
detects the presence of streaks and smears.
19. The method of claim 1, wherein said analyzing step produces a
plurality of physical measurements of said digital pattern
data.
20. The method of claim 19, further comprising: combining said
physical measurements into a quality rating.
21. The method of claim 20, wherein said combining step comprises:
producing a plurality of weights, each weight assigned to one said
physical measurement; normalizing said physical measurements;
computing said quality rating by multiplying each said weight by
the respective said normalized physical measurement; and summing
the products of said weights and said measurements into one overall
quality rating.
22. The method of claim 1, wherein said analyzing step identifies a
plurality of factors responsible for quality deviation and suggests
appropriate corrective action.
23. The method of claim 1, wherein said analyzing step uses a
database with historical data unique for the digital printer.
24. The method of claim 1, wherein said analyzing step uses a
database with quality data correlated with printer setting
adjustments.
25. The method of claim 1, wherein said test pattern data includes
samples of horizontal and vertical lines.
26. The method of claim 25, wherein said horizontal and vertical
lines are separated by different distances.
27. The method of claim 1, wherein the test pattern data is stored
as a bitmap data or represented as a page in a page description
language.
28. A computer program, residing on a computer-readable medium,
comprising instructions causing a print analyzer system to: produce
digital test pattern data, said test pattern including one or more
target objects designed to reveal one or more specific printed
image characteristics; send said digital test pattern data to a
digital printer to generate a printed image at said digital
printer; receive digital pattern data generated from scanning said
printed image; and analyze said digital pattern data to generate
one or more quantitative ratings with respect to said one or more
printed image characteristics.
29. The computer program of claim 28, wherein analyzing of said
digital pattern data includes generating a quality rating related
to image quality from said one or more quantitative ratings.
30. The computer program of claim 28, further comprising
instructions causing a print analyzer system to: indicate a setting
adjustment to said digital printer in response to said one or more
quantitative ratings.
31. The computer program of claim 28, further comprising
instructions causing a print analyzer system to: automatically
adjust a setting to said digital printer in response to said one or
more quantitative ratings.
32. The computer program of claim 28, wherein said quantitative
ratings include a physical measurement of said one or more objects
in said digital pattern data.
33. The computer program of claim 28, wherein analyzing of said
digital pattern data produces a plurality of physical measurements
of said digital pattern data and further comprising instructions to
cause a print analyzer system to combine said physical measurements
into a quality rating.
34. The computer program of claim 28, wherein said digital pattern
data has a higher resolution than said printed image.
35. The computer program of claim 28, further comprising
instructions causing a print analyzer system to: read historical
data in a database; and evaluate said one or more quantitative
ratings with said historical data.
36. A print analyzer system for measuring image characteristics of
printed output from a digital printer, comprising: a source of
digital pattern data, said test pattern including one or more
target objects designed to reveal one or more specific printed
image characteristics; a digital printer that receives said digital
pattern data as input and outputs a printed image; a scanner that
receives said printed image as input and outputs digital pattern
data having higher resolution than said printed image; and an image
quality analyzer receiving said digital pattern data and analyzing
said digital pattern data to generate one or more quantitative
ratings with respect to one or more printed image
characteristics.
37. The print analyzer system of claim 36, further comprising: a
database with historical data for said digital printer.
38. The print analyzer system of claim 36, further comprising: a
database with quality data correlated with printer setting
adjustments.
39. A digital printer for measuring image characteristics of
printed output from said digital printer, comprising: a print
engine to receive test pattern data and generate a printed sheet
having a printed image that is delivered to a routing path, said
test pattern including one or more target objects designed to
reveal one or more specific printed image characteristics; and an
integral scanner that interacts with said routing path to scan an
image on said printed sheet and generates digital pattern data,
said scanner having a higher resolution than said print engine.
40. The digital printer of claim 39, further comprising: an
analyzer included in a digital printer controller of said print
engine, said analyzer receiving said digital pattern data and
analyzing said digital pattern data to generate one or more
quantitative ratings with respect to said one or more printed image
characteristics.
41. The digital printer of claim 39, further comprising: an
indicator to indicate unacceptable print quality in response to
said one or more quantitative ratings.
42. The digital printer of claim 39, further comprising: a setting
adjustment indicator to respond to said one or more quantitative
ratings.
43. The digital printer of claim 39, further comprising: an
automatic print setting adjustment in response to said one or more
quantitative ratings.
44. A test instrument for measuring image characteristics of
printed output from a digital printer, comprising: an input that
receives printed pattern data based on scanning an image printed by
a digital printer in response to test pattern data, said test
pattern including one or more target objects designed to reveal one
or more specific printed image characteristics; an analyzer to
receive said digital pattern data and analyze said digital pattern
data to generate one or more quantitative ratings with respect to
said one or more printed image characteristics; and an output
responding to said one or more quantitative ratings.
45. The test instrument of claim 44, wherein said analyzer
identifies factors responsible for quality deviation.
46. The test instrument of claim 44, wherein said analyzer
generates a quality rating related to image quality from said one
or more quantitative ratings.
47. The test instrument of claim 44, further comprising: a setting
adjustment indicator to respond to said one or more quantitative
ratings.
48. The test instrument of claim 44, wherein said output
automatically adjusts a setting to said digital printer.
49. The test instrument of claim 44, wherein said quantitative
ratings include a physical measurement of said one or more objects
in said digital pattern data.
50. The test instrument of claim 44, wherein said analyzer produces
a plurality of physical measurements of said digital pattern data
and combines said physical measurements into a quality rating.
51. The test instrument of claim 44, wherein said analyzer suggests
appropriate corrective action.
52. The test instrument of claim 44, further comprising: a database
having historical data for said digital printer.
53. The test instrument of claim 44, further comprising: a database
having quality data correlated with said digital printer setting
adjustments.
Description
BACKGROUND OF THE INVENTION
[0001] This invention relates to measuring image characteristics of
printed output from a digital printer.
[0002] A digital printer receives electronic digital input data in
the form of a bitmap (in which there is one bit per pixel), and
outputs a printed sheet. The input data are typically received
serially in a raster pattern and are used to control the deposit of
toner from a drum to a sheet of paper.
[0003] Many factors can cause the print quality of a digital
printer to deteriorate over time. Traditionally, assessments of
this degradation have been done subjectively, as a user tries to
match the printer's image quality with his or her subjective
perception of ideal output.
[0004] Generally, printed sheets are manually inspected to
determine quality. The user generates a test page and judges its
quality by mentally recalling the appearance of the test page when
the printer was new or last serviced. The user can also compare the
test page to samples in a "limits" book, which defines the lower
bounds for standard printer output. An actual printed image may
have defects such as white pinholes in black areas, smears, or
blurry edges. Some defects may be acceptable and some may fall
below an acceptable quality level. Based on manual inspection, the
user rejects printed sheets or makes adjustments to the
printer.
[0005] A technician performing preventive maintenance may obtain
readings of the print density uniformity by using a densitometer,
which measures light reflected from a page. These readings only
partially correlate with human judgments of print quality.
Therefore, unless there is a dramatic shift in quality, or a
noticeable defect such as streaking, users have difficulty
detecting subtle or gradual changes in quality. With color output,
detecting changes is even more difficult since the eye is poor at
identifying absolute colors.
SUMMARY OF THE INVENTION
[0006] In one aspect, the invention features, in general, measuring
image characteristics of printed output from a digital printer by
sending test pattern data to the digital printer, generating a
printed image of the test pattern data at the digital printer,
scanning the printed image to obtain digital pattern data, and
analyzing the digital pattern data of-the printed image. The test
pattern that is printed includes target objects designed to reveal
specific printed image characteristics, and the analysis of the
data from scanning the printed image includes the generation of one
or more quantitative ratings with respect to printed image
characteristics.
[0007] In other aspects, the invention features computer programs,
print systems, digital printers, test instruments, and services
that carry out the method of measuring print image characteristics
just described.
[0008] Embodiments of the invention may include one or more of the
following features. The quantitative ratings can include
measurements of the black and white densities, uniformity, edge
sharpness, resolution, and positional accuracy of the target
objects. The measurements also detect defects, such as streaks and
smears. Landmarks in the test pattern are used to correct for
scanner error.
[0009] One or more quality ratings can be generated from the
measurements. Using these ratings, one can automatically monitor
and adjust the print image characteristics of a digital printer so
that the printer is within allowable margins. For example, one can
use the measurements to adjust print engine related parameters such
as power and duration of the laser, and charging current. The
method is applicable to black and white as well as color printing
applications.
[0010] Other aspects and advantages of the invention will be
apparent from the drawings taken together with the accompanying
description.
BRIEF DESCRIPTION OF THE DRAWINGS
[0011] FIG. 1 is a diagram of a print analyzer system in accordance
with the invention;
[0012] FIG. 2 is an illustration of test pattern data;
[0013] FIG. 3A is a diagram of the print quality analyzer
components;
[0014] FIG. 3B is a diagram outlining the series of tests conducted
by the print analyzer system of FIG. 1;
[0015] FIG. 4A is an illustration of pixel clusters;
[0016] FIG. 4B is a flow chart of the print density uniformity
metric;
[0017] FIG. 4C is a representation of pixel density across a target
illustrating a streak condition;
[0018] FIG. 4D is a representation of pixel density across a target
illustrating a smear condition;
[0019] FIG. 5 is a diagram of a printer having a built-in scanner;
and
[0020] FIG. 6 is a diagram of a computer system having a quality
analyzer.
DETAILED DESCRIPTION
[0021] Referring to FIG. 1, print analyzer system 10 is shown.
Bitmap 12 represents a raster image for a test pattern that is
input to a black and white digital printer 14 being evaluated. Each
bit defines whether a corresponding pixel is black or white. The
output of digital printer 14 is printed image 16 on a sheet of
paper. Bitmap 12 may be produced by scanning a printed test pattern
to generate a raster image, may be generated electronically in the
first instance, or may be based on a combination of inputs.
[0022] To measure the image characteristics of output from digital
printer 14, printed image 16 is scanned by scanner 18, and the
output from scanner 18, namely electronic digital pattern data 20,
is analyzed for print image characteristics at print image analyzer
22, which is implemented in a personal computer. Scanning is done
by feeding printed image 16 into a high-resolution scanner or by
having digital printer 14 route printed image 16 directly to a
built-in scanner (see FIG. 5). Scanner 18 does not have to operate
at the rated printer speed since it is only used to perform image
measurements. However, scanner 18 must have sufficient spatial
resolution to capture the details in printed image 16. In general,
scanner 18 should have at least twice the resolution, both
horizontally and vertically, as printer 14 being evaluated. Digital
printer 14 can also route a test page to scanner 18 during normal
high speed printing operations, thereby testing a printed sheet at
a random or pre-defined interval in a printer's output cycle.
[0023] Scanner 18 outputs digital pattern data 20, which is a high
resolution digital representation of printed image 16. Digital
pattern data 20 is a rectangular array of numerical values stored
in memory. There are 8 bits per scanner pixel; these numerical
values range from 0, which denotes pure black, to 255, which
denotes pure white. The data for the image on printed image 16 are
stored in a buffer thereafter accessed by print image analyzer 22.
Values between 0 and 255 are variations from pure black to pure
white. Print image analyzer 22 evaluates the print image
characteristics of the scanned image by conducting a series of
tests on digital pattern data 20. Based on the tests, print image
analyzer 22 generates quality ratings and/or control signals to
make adjustments to digital printer 14. For example, print image
analyzer 22 can indicate that adjustments should be made to bias
voltages, surface charge levels, laser exposure levels, and printer
speeds in digital printer 14. If an adjustment can be made using
software (e.g., by setting parameters in control memory for digital
printer 14), print image analyzer 22 can send commands to adjust
the appropriate parameters, measure the resulting quality, and then
iteratively adjust them.
[0024] Referring to FIG. 2, test image 16 includes target objects
designed to reveal specific imaging characteristics of digital
printer 14. The upper portion 15 of printed image 16 contains
samples of different text, graphics, and image patterns and is
intended for visual quality ranking (not shown). Approximately the
lower third of printed image 16 contains target objects 16a-f that
are analyzed without human intervention. Target objects 16a-f are
test patterns designed to indicate image defects and are analyzed
to evaluate print image characteristics. They are shown
diagrammatically on FIG. 2 and described in more detail below. In
the corners of printed image 16 are landmarks 26a, 26b, 26c. Before
evaluating the image, print image analyzer 22 has a landmark
registration process to compensate for imperfect scanners and to
identify the expected locations in the buffer for target objects
16a-f.
[0025] To compensate for imperfect scanners, the print analyzer
system aligns digital pattern data 20 with a template to account
for any skewing or margin deviations between digital printer 14 and
scanner 18. Print image analyzer 22 stores a template in memory
that represents the ideal location of pixels within landmark 26a
and uses a signal processing/statistical process that employs a
cross correlation function to locate the first landmark 26a. The
cross correlation function sums the product of each pixel within
the template and a corresponding pixel within a search area in
digital pattern data 20 where landmark 26a is expected to be
located. The search process is iterative and correlates the
template with a number of x,y locations for the stored digital
pattern data. The correlation between a template T placed at (p,q)
within the scanned image S is computed as
C(p,q)=.SIGMA.T(x,y)*S(x-p, y-q)/K
[0026] for all x,y within the template, and where the normalizing
coefficient K is computed as
K=.SIGMA.T(x,y)*T(x,y)
[0027] for all values of x,y within the template. This is done when
the template is placed at different p, q values. The range of p, q
values to use is determined from a setup file.
[0028] The highest correlation is obtained when the template is
positioned over landmark 26a, thus determining the position of the
first landmark 26a. The position (p,q) where C(p,q) is maximized
for a given search area is used for the midpoint of landmark 26a.
After finding the first landmark 26a, landmark 26b and landmark 26c
are located in a similar manner.
[0029] Accounting for three landmarks, rather than only two, allows
linear variations in the X and Y directions of scanner 18 to be
corrected. Print image analyzer 22 measures the distances between
landmarks 26a, 26b, and 26c, compares the actual distances with the
ideal distances, and calculates X and Y correction factors. Using
these correction factors, digital pattern data 20 is "stretched" or
"warped." For example, if the ideal distance between landmark 26a
and landmark 26b were 6000 pixels and the actual measurement was
5900 pixels, a linear scaling in Y of 6000/5900 would correct the
scanning error. Thus, using the formula Y'=Y * Correction would
account for scanner slippage in Y.
[0030] After correcting digital pattern data 20 for scanner error,
individual target objects 16a-f are located using relative
positioning techniques. White space that surrounds the target
objects and that is not part of target objects 16a-f is removed by
dividing digital pattern data 20 into smaller images, one image per
target object 16a-f. This technique of isolating target objects
provides a more efficient use of memory because pixels representing
white space are not stored.
[0031] Relevant physical measurements (referred to as "metrics"
herein) known to affect print image characteristics are calculated
for each target object 16a-f. Computing a density distribution for
a given target object and locating the boundaries of a target
object are common techniques used by several image characteristic
measurements carried out by print image analyzer 22. These
techniques will be described first.
[0032] A number of metrics require the density distribution for a
respective target object 16a-f. The density distribution can be
depicted as a histogram showing a density range from 0 through 255
along the X axis and the number of pixels within each density range
measured along the Y axis. Generally, values between 0 and 127 are
black, and values between 128 and 255 are white. Pb(d) and Pw(d)
represent the frequency of black and white pixels, respectively, in
a target object. Values for Pb(d) and Pw(d) range from 0.0 to 1.0.
The means M.sub.b and M.sub.w of the black and white distributions
are the average black and white densities within a black and white
target. M.sub.b and M.sub.w are computed as
M.sub.b=.SIGMA.d*Pb(d), for d=0 to 255
M.sub.w=.SIGMA.d(Pw(d), for d=0 to 255
[0033] Variances V.sub.b and V.sub.w, on a scale of 0-255 , of the
black and white distributions represent how close the densities are
to M.sub.b and M.sub.w within a black or white target.
[0034] V.sub.b and V.sub.w are computed as
V.sub.b=.SIGMA.d.sup.2*Pw(d)-M.sub.b*M.sub.b, for d=0 to 255
V.sub.w=.SIGMA.d.sup.2*Pw(d)-M.sub.w*M.sub.w, for d=0 to 255
[0035] A number of metrics require determining the location of
boundary pixels for a target object and analyzing the pixels near
the boundary. An adaptive thresholding technique identifies
foreground and background pixels in a target object. The technique
establishes a density range for the foreground and background
colors by first computing a density distribution for the isolated
image. Then, the mean and the variance for the black and white
values is calculated. Densities within 2 or 3 standard deviations
(calculated by taking the square root of the variance), of the
black mean are considered foreground values. After knowing which
values are foreground and background values, the target object is
traced and the boundary of the target object is stored as
coordinate pairs.
[0036] The technique that traces the boundary is known as the
Ledley bug follower and described, for example, in Pratt, W.
Digital Image Processing, (John Wiley & Sons, N.Y. 1978) which
is hereby incorporated by reference. This technique locates a
starting position on the border of an object and traces the border
using a column by column, row by row search for foreground pixels
until the original starting position is located again. A foreground
pixel is located by finding the first pixel that follows three
background pixels in a given row. After locating a pixel in the
boundary of an object, the pixel is stored as an X,Y coordinate
pair, and the technique checks each of the eight neighboring pixels
in a counterclockwise fashion for the next boundary pixel,
beginning with the pixel immediately to the right of the boundary
pixel located. If a new boundary pixel is detected, that pixel's
position is marked and the search begins again in the same manner.
Tracing continues until the technique returns to the initial
starting position.
[0037] Referring now to FIG. 3A, components of print image analyzer
22 are shown. After the landmark registration process 28 completes,
the analysis software computes a set of metrics and, from the
results of the measurements, generates quality ratings and printer
control signals 31, as discussed below. Print image analyzer 22 may
also have a database 32 that stores historical data regarding the
image characteristics of digital printer 14 or of other printers of
the same model. The historical data is a resource for problems that
existed in the past. For example, the historical data may include a
table that lists problems in one column and solutions to the
respective problem in a second column. A solution can indicate a
setting adjustment of digital printer 14. Historical data can also
include image characteristics and quality ratings from prior test
results that are compared to the current image characteristics and
print quality ratings to determine how the quality has degraded
over time. Print image analyzer 22 may also have a user interface
34 so that a human can interact with the analysis software.
[0038] Metric Computation
[0039] Referring to FIG. 3B, the metrics computed by the analysis
software of the print image analyzer are shown. First, print image
analyzer 22 computes the print density uniformity metrics (step
36), which measure variations in the print density for the
foreground and the background of the vertical and horizontal solid
black target objects 16a, 16b (shown in FIG. 2) and white target
objects 16c, 16d (i.e., white space directly to the left and
directly above black target objects 16a, 16b respectively). Print
density uniformity metrics determine the amount of texture in
target objects 16a, 16b, 16c, 16d. The print density uniformity
metrics calculated are listed in Table 1.
1TABLE 1 Print Density Uniformity Metrics Average Cluster Size
Variance of Cluster Size Number and Size of Defects
[0040] Referring now to FIG. 4A, clusters of pixels are
illustrated. A cluster is a group of adjacent pixels having a
density that falls within a given range. Ideally, there is only one
cluster for a given target object that was intended to be printed
as a uniform density object. Typically however, a target object has
several pixel clusters that convey texture to the target object.
For example, the density range of pixels in clusters 50 and 51
differ from the density range of pixels in cluster 52 and cluster
53.
[0041] Referring now to FIG. 4B, steps used to compute print
density uniformity metrics (step 36) are shown. A boundary trace is
performed to locate the target object (step 56) and a print density
range between 0 and 255 is selected (step 58). Within the target
object, clusters of pixels having the selected density range are
counted (step 60) using a recursive process. The recursive process
examines the eight neighboring pixels surrounding a current pixel
and marks the neighboring pixel as belonging to the cluster if the
neighboring pixel falls within the selected distribution range. The
count is stored in memory (step 62). This is done for different
density ranges to identify all clusters within the target
objects.
[0042] The print density uniformity computations then measure the
average cluster sizes (i.e., area), the variance of cluster sizes
for a range of densities within a selected variation of the mean
foreground density (M.sub.b and M.sub.w, as calculated above), and
the number and sizes of defects. The smaller the cluster sizes, the
less noticeable any texture will be, although one large cluster
would be ideal. The less varied the cluster sizes the less
noticeable the texture will be.
[0043] Clusters having density values that are 25% or more from the
mean density values are considered defects. A white defect count
measures the number and size of voids in a black area. A black
defect count measures the speckles or random black spots found in
the white area.
[0044] Referring back to FIG. 3B, the second set of metrics deals
with- streaks and smears (step 38). These computations detect
streaks and characterize their width, frequency, and density.
Listed in Table 2 are the specific metrics and ratings for
detecting and characterizing streaks and smears.
2TABLE 2 Streaks and Smears Metrics Side to Side Uniformity Top to
Bottom Uniformity Number of Streaks Width of Streaks Presence of
Smearing
[0045] First, to detect vertical streaks and smears a single row of
density values that represents the entire horizontal object 16b is
computed. This row of values is referred to as a horizontal density
profile. A single pixel in the horizontal density profile is
calculated by averaging all values for pixels in the same column of
target object 16b. Likewise, a vertical density profile for
horizontal target object 16a is produced to detect horizontal
streaks and smears. A vertical density profile takes the average of
all pixels in a row to generate a single value in the vertical
profile.
[0046] To measure side-to-side uniformity, a straight line
approximation of the horizontal profile is computed using a linear,
least-squares fitting method. Any value more than 1 or 2 standard
deviations from the mean density is set to the mean density to
minimize the effects of streaks. The slope of the line measures how
uniform the density is from one side of the target object to the
other side of the target object. The density does not vary if the
slope is 0. To measure a top-to-bottom uniformity, the same
calculations are carried out on the vertical profile.
[0047] Referring now to FIG. 4C, the presence of streaks is
detected by cross correlating a Gaussian function 64 that models a
streak density against the profile 66 (either the horizontal or
vertical profile). The result of the cross correlation for
different positions of the model is stored in an array that is
indexed by position. The number of streaks detected and their
perceptible visibility are determined by ranking the correlation
results and finding those with high correlation values. The width
of the streak is established by cross-correlating the profile with
streak models having different widths and finding the best match
(i.e., the highest correlation value).
[0048] Referring now to FIG. 4D, toner smearing is detected by
fitting the horizontal and vertical profile values to a higher
order polynomial equation, such as
D(x)=ax.sup.3+bx.sup.2+cx=d
[0049] where D is the density value, x is the position or column
index within one row, and a, b, c, and d are coefficients of the
approximation. Toner smearing could also be detected by
approximating the profile as a Fourier Series to measure the degree
that the print density deviates from the ideal case of a straight
line. When using a polynomial approximation, the second and third
order coefficients indicate the degree the profile deviates from a
straight line, which would be the first order coefficient. The sum
of the absolute value of the magnitude of all non-linear
coefficients can be used as an indication of how much the profile
deviates from an ideal case.
[0050] Referring back to FIG. 3B, the third set of metrics 40
measure positional accuracy by analyzing the consistency by which
imaged dots on the printed page are placed in the vertical and
horizontal directions. Horizontal positional accuracy is the
difference between an actual horizontal position and the expected
horizontal position for a vertical line. Vertical positional
accuracy is the difference between an actual vertical position and
the expected vertical position for a horizontal line. The vertical
and horizontal lines in target objects 16f are used for this test.
In addition to horizontal and vertical accuracy, which detect
jitter, the metric determines the deviation from 90.degree. between
a horizontal and a vertical line. Table 3 summarizes the specific
metrics and ratings for positional accuracy.
3TABLE 3 Positional Accuracy Metrics Average Deviation from
Horizontal Line Average Deviation from Vertical Line Deviation from
90.degree.
[0051] Positional accuracy metrics also use target objects 16e,
which are groups of solid black squares having varying sizes and
spaces between squares. Target objects 16e are used to determine
jitter and skew.
[0052] In calculating the positional accuracy metrics, a boundary
trace of the target object (using the medley bug follower previous
described) is first carried out. To determine horizontal accuracy,
a measurement of the difference in the vertical position of a
horizontal line over several rows is taken. Vertical accuracy is
measured by taking the difference in the horizon position of a
vertical line over several columns. Another method for determining
vertical and horizontal jitter generates straight line
approximations, using a minimum least squares fitting technique,
for positions in square target objects and computes the average
vertical and horizontal deviation. For a given straight line
approximation, the normal distance between each edge point and a
straight line is computed using a vector cross-product.
[0053] The amount by which adjacent edges of the target deviate
from 90 degrees is also measured to determine the amount of skew.
The formula that determines the amount of skew uses simple
trigonometry and the slopes of the straight line approximations for
two adjacent sides of the target object. The equation that
determines the angle between the line approximation and the x-axis
is
.theta.=arctan (.DELTA.Y/.DELTA.X)
[0054] Referring once more to FIG. 3B, the fourth set of metrics
measure edge sharpness (step 42). Table 4 summarizes the specific
metrics and ratings for edge sharpness.
4TABLE 4 Edge Sharpness Metrics Edge Uniformity Edge Blur
[0055] Referring to FIG. 2, edge sharpness metrics use target
objects 16e, which are groups of solid block squares having varying
sizes and spaces between squares. The blocks vary in size from
1.times.1 pixel to 60.times.60 pixels.
[0056] Edge sharpness is computed by examining an edge density
profile for pixels immediately outside a boundary and computing how
close the profile approximates an ideal edge. An edge profile is
obtained by averaging the print density of all pixels located one
pixel from the boundary, then averaging the print density of all
pixels located two pixels from the boundary, and so on up to a
predetermined number such as five pixels from the boundary. Each
average becomes one value in the profile. From the edge profile,
metrics for edge uniformity and edge blur are computed.
[0057] Edge uniformity measures the variance of print density when
transitioning from the foreground to the background. The
measurement sums the variances in the edge profile values. The
smaller the sum of the variances is for a given range of pixels,
the sharper the edge appears, whereas the larger the sum, the more
ragged and blurred the edge appears. A weighted sum of the
variances can be used to accentuate the importance of pixels close
to the edge versus those farther from the edge.
[0058] Edge blur measures how close the edge profile approximates
the ideal edge. The ideal edge profile transitions from solid black
to white in one or two pixels, almost in a straight-line fashion.
Edge blur approximates the edge profile as a straight line and uses
the slope of the line approximation as an indication of blur, where
the higher the slope the sharper the image and the lower the slope,
the longer the transition from black to white and the more blurred
the edge appears.
[0059] Referring once more to FIG. 3B, the fifth metric computes
edge acuity (step 44). Edge acuity measures the ability of digital
printer 14 to reproduce detail. The metric has only one component,
as shown in Table 5.
5TABLE 5 Edge Acuity Metric Edge Acuity
[0060] Referring to FIG. 2, target objects 16f are horizontal,
vertical, and 45.degree. lines used for determining edge acuity.
The edge acuity metric finds the closest separable line pairs
within target objects having pairs of lines separated by varying
distances. The spacing between the lines are increased by one
printer unit for every line pair. The thickness of the lines can be
held constant or varied. To account for imaging capabilities in the
horizontal and vertical directions, the test uses three sets of
target objects, one with vertical lines, one with horizontal lines,
and one with 45.degree. lines.
[0061] The edge acuity metric (step 44) locates the lines and
computes the distance between the opposing edges of the closest
lines. If a separation between the lines is detected, the distance
is considered resolvable. When the line pair distances drops below
the resolvable capability of the printer, all lines fuse into one
object and are considered as one object. This is detected because
the size of the identified object is greater than the size of the
target (i.e., the identified object would include two or more
targets).
[0062] Quality Ratings and Printer/Control Signal Generation
[0063] Quality ratings and printer control signals are generated
from the metrics in process 31 (FIG. 3A). The quality ratings
include an overall quality rating and more specific quality
ratings, for example with respect to print density uniformity,
streaks, smears, positional accuracy, edge sharpness, and edge
acuity. To generate the various quality ratings, first the metric
values in Tables 1-5 are normalized to values between 0.0 and 1.0
and the weights are applied to the normalized values. For example,
a lower natural limit for a metric would be 0.0, and an upper
natural limit would be 1.0.
[0064] A weight is assigned to each physical measurement listed in
Tables 1-5 and defines the relative significance of each
measurement to the respective quality rating. The weight is
multiplied by the normalized quality rating of the respective test.
The overall quality rating is the sum of all the weighted and
normalized quality ratings for the individual tests.
[0065] Using the quality ratings, print image analyzer 22 can
identify factors responsible for a quality deviation and suggest an
appropriate corrective action to tune and calibrate digital printer
14 to an acceptable quality level. For example, an unacceptable
rating for the presence of smears may indicate that the toner
cartridge needs replacing. Furthermore, ratings and weights given
to the measurements can depend on the age and model of the digital
printing device.
[0066] Other embodiments are within the scope of the following
claims. For example, print image analyzer 22 can be part of the
controller firmware of digital printer 14. This embodiment enables
digital printer 14 to directly measure its own output quality. FIG.
5 shows scanner 18 built into the output path of digital printer 14
and receiving test pages as needed from printing drum 15. Each time
digital printer 14 runs diagnostics, printed image 16 would be
printed, scanned, and compared with previous measurements. If the
print quality degraded below a threshold or a serious problem were
detected, the operator would be alerted by an alarm or light
indicator.
[0067] In an alternative embodiment, such as shown in FIG. 6, print
image analyzer 22 can be integrated into an image quality assurance
system for printers in a distributed system. The results of
performing the analysis could be stored locally with each printer
or centrally in an administrative site.
[0068] In yet another embodiment, print image analyzer 22 can be
integrated into a service kit used by a technician to measure image
characteristic deviations and determine appropriate corrections. A
database could track problems and corrections that correlate with
different types of measurements taken from the test sample. A left
edge smudge, for example, might correlate with a misaligned toner
roller. The initial scan of the printed page provides the
technician with a list of potential problems and likely
corrections. Once the problem is identified and corrected, the
technician updates the database with the information.
[0069] Print image analyzer 22 can be integrated into a printer's
test fixture to provide continuous on-line measurements of print
image characteristics. This would be especially useful for high
volume printers that require large volumes of output to show
significant quality changes. The ability to continuously sample the
print output, plot variations, and correlate those results enables
a printer manufacturer to predict quality degradation and thus
reduce testing costs.
[0070] Print image analyzer 22 can also be used to provide on-line
metrics for a printer manufacturer's quality assurance or research
and development departments.
[0071] FIG. 6 shows a computer system 100 suitable for supporting
one or more print image analyzers 102. The computer system 100
includes a digital computer 104, a display monitor 106, a keyboard
108, a mouse or other pointing device 110, and a mass storage
device 112 (e.g., hard disk drive, magneto-optical disk drive, or
floppy disk drive). The computer 104 includes memory 120, a
processor 122, and other customary components, such as, memory bus
and is peripheral bus (not shown). The computer 104 has a network
interface 124 to communicate with remote computer systems. Digital
printer 14 and scanner 18 also connect to network interface
124.
[0072] In place of bitmaps, the test pattern data can be in the
form of page description language such as PostScript or HP's
printer control language (PCL).
* * * * *