Television – Special applications – Film – disc or card scanning
Reexamination Certificate
1999-03-16
2004-08-24
Vo, Tung T. (Department: 2712)
Television
Special applications
Film, disc or card scanning
C358S506000
Reexamination Certificate
active
06781620
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates generally to digital film developing/processing and more particularly, to apparatus and methods for timed-scan image stitching and noise reduction.
BACKGROUND OF THE INVENTION
Digital film processing (“DFP”) promises several advantages over traditional “darkroom” film development techniques, including more accurate image reproduction, defect correction, and image processing, among others
A DFP system for capturing an image recorded on photographic film is illustrated in
FIG. 1
, which figure is taken from FIG. 9 of U.S. Pat. No. 5,519,510. As shown, conventional color photographic film
101
typically includes three film layers, each of which is constructed to record different primary color information (i.e. red, green and blue color information) of a source image during picture taking.
In one aspect of the '510 patented invention, the DFP system captures image information from film
101
by projecting infrared light
11
and
12
at and correspondingly scanning reflected and/or transmitted light from successive portions while the film is being developed with a monochromatic developer. A complete image dataset is formed by capturing successive image portions of successive film layers and then repeating this process at multiple times while film
101
is developed.
For example, scanning a three-layer film
101
includes capturing reflected light
11
from a portion (e.g. portion
11
a
) of a first or “front” film layer
111
and capturing reflected light
12
from a corresponding portion (e.g. portion
12
a
) in a third or “back” film layer
113
. Such scanning also includes capturing a corresponding portion (e.g. portion
11
a
) in the second (i.e. “middle” or “through”) layer by scanning transmitted light
11
passing through film
101
, and then subtracting scanned front layer and back layer values for corresponding front layer and back layer portion scans. A complete image dataset therefore includes front-layer, through-layer and back-layer image pixel data for each portion of the recorded image (in each film layer) at each scan-time during film development.
As shown in
FIGS. 2A through 3
, multiple scans at increasing times during film development (“timed-scans”) are used to enable more accurate reproduction of a source image. Beginning with
FIG. 2A
, during picture taking with camera
210
, a single image
201
recorded onto film
101
(
FIG. 1
) will typically include discernable picture elements such as highlights
211
a
, midtones
21
b
and shadows
211
c
. Turning to
FIG. 2B
, during development of film
101
, an early scan
202
a
(e.g. one minute) will best reveal highlights
211
a
, while midtones
211
b
and shadows
211
c
will be under-developed. A laterscan (e.g. two minutes) will better reveal midtones
211
b
, while highlights
211
a
will become overdeveloped. Still later scans will better reveal shadows
211
c
at the expense of highlights
211
a
and midtones
211
b
. Thus, an image dataset comprising multiple timed-scans representing image elements (e.g. highlights, midtones and shadows covering a complete development time range is used to enable an accurate exposure of each element to be determined.
As shown in
FIG. 3
, a sufficient number of scans are taken such that the exposure of selected picture elements in each film layer can be deduced. Individual scans can further be combined for reducing memory utilization. (Scans
302
and
304
, for example, can be combined to produce scan
308
.)
FIGS. 3 through 5
illustrate examples of systems for combining image data contained in an image dataset (“stitching systems”). A key goal of such systems is to form, from the timed-scans in an image dataset, a resultant image dataset wherein each image element is represented at the brightness or “exposure” at which it was recorded onto film. For example, in the system of
FIG. 3
(hereinafter referred to as “splice-stitching”) an approximation is made as to the best exposure for each group of picture elements utilized (i.e. in this case, highlights, midtones and shadows). In addition, the groups are aligned, cut and pasted to form a single image. Unfortunately, approximating exposure values (which might be between available timed-scan data), such that each element is properly exposed and a visually continuous single image is formed, has been found to be computationally expensive and to yield inconsistent results.
FIG. 4
illustrates an alternative stitching system, referred to as “parametric-stitching,” taught by a U.S. Patent Application entitled “Parametric Image Stitching” filed Feb. 22, 1999 and based upon Provisional Application No. 60/075,562, which application is assigned to the same assignee as that of the present invention. Captured pixels within an image dataset are preferably used as image elements for stitching purposes. As described, timed-scan pixel data is tagged with a capture-time (i.e. a relative and/or absolute indicator of the time of each scan during film developing). Further, for each pixel at each scan-time, regression parameters are calculated and. summed, the optimal density of each pixel position on the film is predicted, a gamma correction function is applied, and a brightness value for each pixel is determined.
FIG. 4
, for example, shows how different curves representing low, medium and high exposure timed-scans are preferably obtained for each pixel based upon a “best fit” of received pixel data for each of the different timed-scans, and an optimum density curve is empirically derived, as shown by dotted line
404
. The actual best “brightness” (or “exposure”) for a pixel is preferably determined based upon the intersection of the optimum density curve with the best fit curve corresponding.to the pixel type (i.e. whether it was a low, mid or high exposure pixel).
Among the advantages of parametric-stitching over splice-stitching is the replacement of “pasting groups of picture elements together” with a more reliable model-based implementation (i.e. where the proper exposure of a pixel can be resolved without reference to the image as a whole)
FIG. 5
illustrates a still further image stitching system as taught by U.S. patent application Ser. No. 09/196,208 filed Nov. 20, 1999 entitled “Log-Time Processing and Stitching System” and is taken from FIG. 5A of that application. As with parametric-stitching, captured pixels within an image dataset are preferably used as image elements for stitching purposes.
In one aspect of this system (hereinafter referred to as “log-time stitching”), a regression analysis is performed that compares image data at various development times versus the natural log of time to obtain a best fit line of this data. The best-fit line is then used to determine a “b” value or “fitting constant” which preferably corresponds to the y-intercept of the best-fit line. It is discovered that this “b” value is substantially directly proportional to the log exposure of a corresponding pixel. As is taught in another aspect of log-time stitching, a “b” value can be calculated using the principles of matrix algebra according to the following equation-1
[
N
∑
ln
(
t
)
∑
ln
(
t
)
∑
ln
(
t
)
2
]
A
[
b
m
]
B
=
[
∑
S
∑
S
(
ln
(
t
)
)
]
C
Equation 1
wherein “N” represents a number of timed-scans made for a given film pixel (or given pixel of a single film layer in a multiple layer film), “t” represents each scan time relative to the start of film developing, “S” represents a detected signal value corresponding to the signal received from the sensor, “m” represents the slope of the best fit line, and “b,” which is referred to as the “b value,” is the y-intercept of the best fit line. Scan times can, as appropriate, be approximated as being equal for all pixels, such that matrix-A of equation-1 can be calculated once and then utilized for determining a “b” value for all pixels.
U.S. Pat. No. 5,519,510 and the Parametric Image Stitching application base upon U.S. Application Nos. 60/075,562 and U.S. application No. 09/196,
Eastman Kodak Company
Galasso Raymond M.
Novais David A.
Vo Tung T.
LandOfFree
Mixed-element stitching and noise reduction system does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Mixed-element stitching and noise reduction system, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Mixed-element stitching and noise reduction system will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3353790