 |
Claims  |
|
|
What is claimed is:
1. A method for operating an image-presenting mechanism responsive to an
application thereto of electrical command signals to display an image that
the command signals represent, the method comprising the steps of:
A) performing a sequence of image-revision steps, in which sequence each
image-revision step receives electrical input signals representing an
input image consisting of input pixels that the input signals represent as
respective input-pixel values and produces therefrom electrical output
signals representing an output image consisting of output pixels that the
output signals represent as respective output-pixel values, the input
image of the first image-revision step is a source image, the input image
of any subsequent image-revision step is the output image of the preceding
image-revision step, and in which:
i) one said image-revision step is a half-toning step whose input image is
input signals represent in a relatively fine-resolution pixel values and
whose output image is output signals representing relatively
coarse-resolution pixel values; and
ii) one said image-revision step, which precedes the half-toning step, is
an infinite-impulse-response-filter step, the values of each of a
plurality of output pixels, for at least some input images, is a filter
function of the value of an input pixel corresponding to that output pixel
and of the values of a respective plurality of the
infinite-impulse-response-filter step's other output pixels; and
B) applying to the image-presenting mechanism the electrical command
signals that represent the output image produced by the last
image-revision step.
2. A method as defined in claim 1 wherein the filter function is a
high-pass-filter function for at least some output pixels.
3. A method as defined in claim 2 wherein the filter function for any given
output pixel is a high-pass-filter function if the magnitude of the
difference between the value of the corresponding input pixel and a
weighted sum of the values of the output pixels in a predetermined
neighborhood of the given output pixel exceeds a predetermined high
threshold.
4. A method as defined in claim 3 wherein the filter function for any given
output pixel is a low-pass-filter function if the magnitude of the
difference between the value of the corresponding input pixel and a
weighted sum of the values of the output pixels in a predetermined
neighborhood of the given output pixel is less than a predetermined low
threshold.
5. A method as defined in claim 4 wherein the high and low thresholds are
equal.
6. A method as defined in claim 1 wherein the filter function is a
low-pass-filter function for at least some output pixels.
7. A method as defined in claim 6 wherein the filter function for any given
output pixel is a low-pass-filter function if the magnitude of the
difference between the value of the corresponding input pixel and a
weighted sum of the values of the output pixels in a predetermined
neighborhood of the given output pixel is less than a predetermined low
threshold.
8. A method as defined in claim 6 wherein one said image-revision step is a
second half-toning step, which precedes the
infinite-impulse-response-filter step.
9. A method as defined in claim 1 wherein one said image-revision step is a
second half-toning step, which precedes the
infinite-impulse-response-filter step.
10. A storage medium containing instructions readable by a computer to
configure the computer to function as an apparatus for operating an
image-presenting mechanism responsive to an application thereto of
electrical command signals to display an image that the command signals
represent, the apparatus comprises:
A) An image-revision circuitry, responsive to electrical source-image
signals representing a source image, for performing a sequence of
image-revision steps, in which sequence each image-revision step receives
electrical input signals representing an input image consisting of input
pixels that the input signals represent as respective input-pixel values
and produces therefrom electrical output signals representing an output
image consisting of output pixels that the output signals represent as
respective output-pixel values, the input image of the first
image-revision step is the source image, the input image of any subsequent
image-revision step is the output image of the preceding image-revision
step, and in which:
i) one said image-revision step is a half-toning step whose input image is
input signals representing relatively fine-resolution pixel values and
whose output image is output signals representing relatively
coarse-resolution pixel values; and
ii) one said image-revision step, which precedes the half-toning step, is
an infinite-impulse-response-filter step, the values of each of a
plurality of output pixels, for at least some input images, is a filter
function of the value of an input pixel corresponding to that output pixel
and of the values of a respective plurality of the
infinite-impulse-response-filter step's other output pixels; and
B) An output circuitry responsive to the image-revision circuitry for
applying to the image-presenting mechanism the electrical command signals
that represent the output image produced by the last image-revision step.
11. A storage medium as defined in claim 10 wherein the filter function is
a high-pass-filter function for at least some output pixels.
12. A storage medium as defined in claim 11 wherein the filter function for
any given output pixel is a high-pass-filter function if the magnitude of
the difference between the value of the corresponding input pixel and a
weighted sum of the values of the output pixels in a predetermined
neighborhood of the given output pixel exceeds a predetermined high
threshold.
13. A storage medium as defined in claim 12 wherein the filter function for
any given output pixel is a low-pass-filter function if the magnitude of
the difference between the value of the corresponding input pixel and a
weighted sum of the values of the output pixels in a predetermined
neighborhood of the given output pixel is less than a predetermined low
threshold.
14. A storage medium as defined in claim 13 wherein the high and low
thresholds are equal.
15. A storage medium as defined in claim 10 wherein the filter function is
a low-pass-filter function for at least some output pixels.
16. A storage medium as defined in claim 15 wherein the filter function for
any given output pixel is a low-pass-filter function if the magnitude of
the difference between the value of the corresponding input pixel and a
weighted sum of the values of the output pixels in a predetermined
neighborhood of the given output pixel is less than a predetermined low
threshold.
17. A storage medium as defined in claim 15 wherein one said image-revision
step is a second half-toning step, which precedes the
infinite-impulse-response-filter step.
18. A storage medium as defined in claim 10 wherein one said image-revision
step is a second half-toning step, which precedes the
infinite-impulse-response-filter step.
19. An apparatus for operating an image-presenting mechanism responsive to
the application thereto of electrical command signals to display an image
that the command signals represent, said apparatus comprising:
A) An image-revision circuitry, responsive to electrical source-image
signals representing a source image, for performing a sequence of
image-revision steps, in which sequence each image-revision step receives
electrical input signals representing an input image consisting of input
pixels that the input signals represent as respective input-pixel values
and produces therefrom electrical output signals representing an output
image consisting of output pixels that the output signals represent as
respective output-pixel values, the input image of the first
image-revision step is the source image, the input image of any subsequent
image-revision step is the output image of the preceding image-revision
step, and in which:
i) one said image-revision step is a half-toning step whose input image is
input signals representing relatively fine-resolution pixel values and
whose output image is output signals representing relatively
coarse-resolution pixel values; and
ii) one said image-revision step, which precedes the half-toning step, is
an infinite-impulse-response-filter step, the values of each of a
plurality of output pixels, for at least some input images, is a filter
function of the value of an input pixel corresponding to that output pixel
and of the values of a respective plurality of the
infinite-impulse-response-filter step's other output pixels; and
B) An output circuitry responsive to the image-revision circuitry for
applying to the image-presenting mechanism the electrical command signals
that represent the output image produced by the last image-revision step.
20. An apparatus as defined in claim 19 wherein the filter function is a
low-pass-filter function for at least some output pixels.
21. An image-presentation system comprising:
A) An image-revision circuitry, responsive to electrical source-image
signals representing a source image, for performing a sequence of
image-revision steps, in which sequence each image-revision step receives
electrical input signals representing an input image consisting of input
pixels that the input signals represent as respective input-pixel values
and produces therefrom electrical output signals representing an output
image consisting of output pixels that the output signals represent as
respective output-pixel values, the input image of the first
image-revision step is the source image, the input image of any subsequent
image-revision step is the output image of the preceding image-revision
step, and in which:
i) one said image-revision step is a half-toning step whose input image is
input signals representing relatively fine-resolution pixel values and
whose output image is output signals representing relatively
coarse-resolution pixel values; and
ii) one said image-revision step, which precedes the half-toning step, is
an infinite-impulse-response-filter step, the values of each of a
plurality of output pixels, for at least some input images, is a filter
function of the value of an input pixel corresponding to that output pixel
and of the values of a respective plurality of the
infinite-impulse-response-filter step's other output pixels;
B) an image-presenting mechanism responsive to the application thereto of
electrical command signals to present an image that the command signals
represent; and
C) An output circuitry responsive to the image-revision circuitry for
applying to the image-presenting mechanism the electrical command signals
that represent the output image produced by the last image-revision step. |
|
 |
|
 |
Claims |
|
|
|
Description |
|
|
BACKGROUND OF THE INVENTION
The present invention concerns image printing. It has particular
application to the image filtering that is employed to enhance feature
rendering in half-toned images.
In most cases a digitally stored image is represented as pixel values that
a device printing the image should attempt to approximate as closely as
possible. But obtaining the desired fidelity is complicated by several
factors. In the case of an ink-jet printer, for instance, the range of
intensities available for a given color component at any given pixel site
is merely binary. That is, an ink dot of the given color component can be
either deposited or not: there typically is no choice in the size of the
ink dot that can be deposited. Therefore, the greater value choice that
the stored image's finer value quantization affords is simulated by
half-toning, in which the finer-resolution value intended at a particular
pixel determines the probability that an ink drop will be deposited at
that pixel. Of course, half-toning affects fidelity adversely.
Additionally, the relationship between the apparent color darkness and the
dot-deposition probability that will produce that darkness is not linear,
so "dot-gain compensation" must be performed on the incoming image in
order to increase fidelity. Similar processing must also be performed to
compensate for the inaccuracy in the relationship between the primary
colors that the stored values represent and the colors of which each ink
actually used is capable.
For these and many other reasons, a lot of image processing must be
performed in order to maximize the fidelity with which the intended image
is rendered. In many cases, this processing involves filtering. Sometimes,
high-pass filters are employed to re-emphasize edge features, which
half-toning and various other processes tend to suppress. Conversely,
low-pass filters are often employed in low-contrast regions in order to
suppress processing artifacts that would otherwise be particularly
apparent in such regions. Indeed, some systems employ both high- and
low-pass filtering the same image: regions identified as being high in
contrast are subjected to high-pass filtering for edge enhancement, while
regions identified as being low in contrast are subjected to low-pass
filtering for smoothing.
When printers that use half-toning employ filtering, they typically use the
same filter types that other image-processing applications use. That is, a
"kernel" of filter co-efficients conceptually overlays a neighborhood of
pixels that are centered on the pixel whose output value is to be
computed, the neighborhood-pixel values are multiplied by the
co-efficients that respectively overlay them, and the output value is the
sum of the resultant products. The filter effect depends both on the
filter co-efficients and on the kernel size. In general, a larger
low-pass-filter kernel is capable of greater smoothing than a smaller
filter kernel.
SUMMARY OF THE INVENTION
We have discovered that the desired filtering of images to be subjected to
half-toning can be obtained at relatively modest computational cost by
employing infinite-impulse-response filters, i.e., filters in which the
filter outputs for at least some pixels are computed not only from filter
inputs but also from the filter outputs computed for those pixels'
neighbors. As is observed in, for instance, Jahne, Digital Image
Processing, 2d Ed., Springer-Verlag 1993, pp. 112-113, such filters had
not heretofore been considered of much use in image processing, because
considerable effort is required to compute the output of a
symmetrical-response filter of that type. But we have found that even
asymmetrical-infinite-impulse-response filters yield beneficial results in
printer applications. Moreover, since the degree of smoothing that can be
obtained for a given kernel size is much greater in an
infinite-impulse-response filter than in a conventional filter, the
computational cost of such a filter is much less.
BRIEF DESCRIPTION OF THE DRAWINGS
The invention description below refers to the accompanying drawings, of
which:
FIG. 1 is a hardware-perspective block diagram of a typical system
employing the present invention's teachings;
FIG. 2 is a block diagram that depicts the same system from a software
standpoint;
FIG. 3 is a block diagram that depicts the processing performed by the
printer driver 36 of FIG. 2;
FIG. 4 is a conceptual diagram of the function performed by the filter
circuit 62 of FIG. 3;
FIG. 5 is a diagram of the filter kernel that the FIG. 4 filter employs;
FIG. 6 is a plot of the single-pixel-disturbance response of a
finite-impulse-response filter whose kernel FIG. 5 depicts; and
FIG. 7 is a plot of the single-pixel-disturbance response of an
infinite-impulse-response filter having the same kernel.
DETAILED DESCRIPTION OF AN ILLUSTRATIVE EMBODIMENT
FIGS. 1-3 will illustrate a typical system in which the present invention's
teachings can be practiced, while its most-characteristic features will be
described by reference to FIGS. 4-7.
As the invention description proceeds, it will become apparent that the
invention can be embodied in dedicated circuitry designed particularly to
implement the invention's teachings. Such an arrangement can be included
within a printer that receives instructions that describe an image in
terms of finely quantized nominal colors or gray-scale values, and it can
include dedicated circuitry that filters the image in the process of
converting the requested values to the on-and-off or other
low-value-resolution instructions required to render the requested image.
But the invention will more typically be implemented by a general-purpose
machine, such as a personal computer operating as a printer driver, whose
purpose is to convert an image expressed in nominal color values into
printer commands that specify the low-level, typically on-or-off operation
of a printer that the computer controls.
FIG. 1 depicts a typical hardware environment. A personal computer 10 sends
an image-presentation device such as an ink-jet printer 12 low-level
instructions, i.e., instructions that specify which individual
print-medium pixels should receive dots. The drawing depicts the printer
12 as receiving these instructions by way of an appropriate channel 14.
Computers that are capable of practicing the present invention come in a
wide variety of configurations, and FIG. 1 depicts one in which channel 14
is provided by an input-output adapter 16 with which a central processing
unit 18 communicates by way of an internal bus 20.
Of course, the central processing unit 18 will typically fetch data and
instructions at various times from a variety of sources, such as
solid-state read-only and random access memories 22 and 24. FIG. 1 also
depicts the computer 10 as communicating, as is typical, with a keyboard
26 by way of an interface adapter 28 and with a cathode-ray-tube display
30 by way of a display adapter 31.
The present invention particularly concerns image-presentation devices
within this environment. Although the computer 10 can in principle employ
the present invention's teachings not only to drive printer 12 but also to
form an image on the cathode-ray-tube display 30, this would not be the
invention's normal use, because display 30's value quantization is
typically much finer--and its spatial resolution much coarser. So we will
restrict our attention to the invention's use for operating the printer
12.
In the typical situation, the computer 10 implements the present
invention's teachings by acting as a printer driver. The instructions that
configure the computer to perform this function are usually contained in
the operating-system software transferred through the computer's disc
drive 32 to a disc 33 for storage. Often, the driver software will have
been loaded into the computer system from another medium such as a
diskette, CD-ROM or DVD. In any event, the computer 10 reads the
printer-driver instructions from the disc drive in most cases and then
performs the below-described functions to implement the present
invention's teachings.
FIG. 2 depicts the typical situation from the software standpoint. The
present invention's teachings typically come into play when the computer
10 is operating a user's application program 34 and that program makes a
system call requesting that an image be printed. The requested operation
is carried out by a printer driver 36, which is usually considered to be
part of the operating system but is specific to the designated printer.
The printer driver's purpose is to convert a device-independent
representation of the image into low-level printer instructions to the
printer that will render that image as faithfully as the printer's
limitations permit.
The filtering operations that particularly distinguish the present
invention's embodiments can be employed in many image-processing
sequences. The sequence that FIG. 3 depicts is typical. A source image of
twenty-four bits per pixel, i.e., eight bits per color component, is
subjected to color-correction operations represented by blocks 52 and 54.
Multi-level dithering 52 quantizes the source values: each color component
in block 52 quantized output can assume one of only seventeen possible
values, which are used to address a color-correction memory in block 54.
The memory's size can therefore be limited to 17.times.17.times.17 instead
of the impractical 256.times.256.times.256 size that would have been
required before multi-level dithering. Commonly assigned U.S. patent
application Ser. No. 08/607,074, filed Feb. 26, 1996, by Shu et al. for
Generating Color-Correction Look-Up-Table Addresses by Multi-Level
Half-Toning, describes these operations further. We incorporate that
application by reference.
The table 54 is used to correct certain imperfections in the printing
process. To signify this, we have included in FIG. 3 two further blocks 58
and 60, which do not represent real-time operations. Instead, they
represent steps that could have been performed in arriving at the
look-up-table contents. These steps may start with look-up-table values
whose purpose is only to increase color-rendition fidelity; these values
simply take into account the non-ideal colors of the inks that various
printers employ. A step represented by block 58 may revise those values to
take into account the effects that ink-dot shapes, which tend to differ
for different media, have on apparent color intensity. We also modify the
look-up-table entries, in a step represented by block 60, to limit ink use
enough to prevent it from bleeding on paper of the type currently being
employed.
In a typical arrangement, each look-up-table location may contain, for
example, four four-bit values, one for each of the four ink colors that
the printer will use: cyan, magenta, yellow, and black. But the present
invention can be practiced in printers and printer drivers--and other
image-presentation mechanisms, for that matter--in which the look-up-table
parameters are quite different. Indeed, the present invention's broader
teachings can be practiced in arrangements that do not employ any look-up
table at all.
For each color component, the signal-processing sequence ends with a binary
half-toning step 61, whose output indicates for each pixel whether that
pixel will receive an ink drop of the type that corresponds to that color
component, and that output is sent to the printer. Although the various
image-processing operations mentioned so far do serve to improve the
printer output's image fidelity, they also tend to introduce artifacts of
their own. So some image-presentation systems include a filtering-type
image-enhancement operation 62 to suppress some of those artifacts and
otherwise enhance the image. FIG. 4 illustrates an example of the type of
filtering that distinguishes the present invention. FIG. 4 depicts the
filtering for one of the components. The other components are filtered
identically.
Although a typical embodiment of the present invention's teachings will
implement the filtering process in software, FIG. 4 depicts a hardware
arrangement for the sake of explanation. As will be explained below, the
FIG. 4 circuit can operate as a low-pass filter or as a high-pass filter,
and it can also simply forward its input without performing any filtering
at all. When it does perform filtering, it computes the filter output for
a given pixel not only from the input value for that pixel and its
neighbors but also, in accordance with the present invention, from
neighboring pixels' output values. This use of output values, rather than
only input values, gives the filter its infinite-impulse-response
character and enables it to provide more smoothing than conventional
filters do for a given kernel size.
A (conceptual) shift register 66 one stage greater in length than the
number of pixels in an image row stores the previous outputs, and
operations represented by multipliers 68 compute the products of those
values and filter-kernel co-efficients of which FIG. 5 sets forth
exemplary values. FIG. 5's dashed-line box 70 represents the position of
the pixel whose output value is being computed, and its solid-line boxes
72 represent the relative positions of the pixels whose previously
determined output values the filter uses to generate that output value.
A summation operation 74 (FIG. 4) computes the sum A of the resultant
products. Another summation operation 76 takes a weighted average of the
subject pixel's input value and the weighted neighborhood output-value
average A. If the filter is to operate in its low-pass mode, the result is
the filter output. The number of possible filter input values is limited
to the relatively small number of look-up-table locations, but the
filtering operation results in a greater number of possible values for the
sum A, and, even though a further half-toning operation follows the filter
operation, we believe that the filter-operation output's finer granularity
tends to suppress some artifacts that the look-up-table operation's
coarser granularity introduces.
To determine whether to use the low-pass function, a comparison operation
80 computes the difference .DELTA. between the average A and the input
value. If that difference's magnitude is lower than a predetermined
low-pass threshold, as determined in a threshold-comparison operation 82,
then the filter operates in its low-pass mode, as an (again, conceptual)
switch 84 indicates.
If .DELTA.'s magnitude exceeds a predetermined high-pass threshold, on the
other hand, the filter computes a high-pass output. In operation 86, it
computes the product of the difference .DELTA. and a proportionality
constant .beta., and it adds the result to the input in operation 88.
(Inspection reveals that the "high-pass" filter output does include a DC
component; i.e., by high-pass we mean a response that emphasizes changes,
even though it may not filter out all low-spatial-frequency components.)
The high- and low-pass thresholds may be the same. If not, one approach is
simply to employ the input as the output if the difference .DELTA. is
between the two threshold values, as three-position switch 84 indicates.
We have discovered that, despite the limitations ordinarily thought to
afflict the infinite-impulse-response filters' use in image processing,
employing them in half-toning applications can yield significant
improvements in image quality, and we have thereby is been able to take
advantage of their modest computational cost. FIGS. 6 and 7 illustrate
that advantage.
FIGS. 6 and 7 depict responses of finite- and infinite-impulse-response
filters, respectively, to a single-pixel disturbance. The two-dimensional
"horizontal" position represents location in the output image, and the
"height" represents the filter output. In other words, the height
represents the output-image darkness in response to an input image whose
pixel value at, say, (4,4) is 255 and zero everywhere else.
In both drawings, the filter kernel is the one that FIG. 5 illustrates. But
any given pixel's output value in the conventional, FIG. 6 case depends
only on the input values of a neighborhood as small as the filter kernel,
so even the small, twenty-pixel-square region that FIG. 6 illustrates
dwarfs the response of a conventional, finite-impulse-response filter. It
is clear that a much larger kernel, and therefore much more computation,
is necessary if any significant smoothing effect is to be obtained. But
the same kernel size yields a much larger response in an
infinite-impulse-response filter, as FIG. 7 shows, because that kernel is
applied to output values rather than input values, so the input values'
effects tend to propagate. Therefore, a given pixel's output value depends
on a large number of input values. Indeed, the output pixel value for the
pixel at the lower right-hand corner theoretically receives contributions
from all input values.
So employing the present invention's teachings provides desired image
enhancement at a computational cost much lower than conventional
approaches impose. It therefore constitutes a significant advance in the
art.
* * * * *
|
|
|
|
|
|
|
Description |
|
