Television – Camera – system and detail – Optics
Reexamination Certificate
1997-08-26
2002-02-26
Garber, Wendy (Department: 2712)
Television
Camera, system and detail
Optics
C348S337000, C348S208400, C348S340000
Reexamination Certificate
active
06351285
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a motion correction device for a camcorder (camera recorder), and more particularly, to the one-dimensional image component pick-up unit for correcting an unstable image from a camcorder caused by unsteady hands.
2. Description of the Related Art
Detection of a motion vector from a dynamic image signal is an essential technique in the compression, recognition, stabilization, etc. of an image. When a portable video camera is used in conjunction with a VCR (videocassette recorder) to take a picture, an image input to the camera is likely to be unstable and shaking, especially when the user is walking or in a moving vehicle. A common problem of camcorders is that when a picture is highly magnified, the instability caused by unstable hands is more pronounced.
A solid state pick-up unit is widely used as an input device of a camcorder. The solid state pick-up unit is a two-dimensional image pick-up unit made from a semiconductor chip which does not use electron beams. There are two types of solid state pick-up units—a MOS type which uses a metal oxide semiconductor (MOS) transistor in a light receiver, and a CCD (charge coupled device) type.
In the following, an existing image correction system will be explained in conjunction with the accompanying drawings.
FIG. 1
is a block diagram showing a conventional image correction system, and
FIG. 2
is a conceptual diagram showing a conventional method of detecting a motion vector for an image by detecting a pair of one-dimensional vector components.
As shown in
FIG. 1
, the conventional image correction system for stabilizing an image comprises a CCD imager
11
(a solid state pick-up unit) which picks up an input image from incident light that has passed through an optical system, scans the image electronically within the solid state device, and converts the image into electrical signals. An analog/digital converter
12
(referred to as an AID converter hereinafter) converts the analog signals output from the CCD imager
11
into digital signals. A camera signal processing unit
13
converts a signal output from the A/D converter into color and brightness signals. A motion vector detecting unit
14
computes a motion vector from the signal output from the A/D converter
12
. A memory control unit
15
receives the motion vector output from the motion vector detecting unit
14
to control the position of pixels in the image. A field memory
16
holds a field unit (or a frame unit) of color and brightness data from the camera signal processing unit
13
and, under the control of the memory control unit
15
, outputs a stabilized image signal. A digital/analog converter
17
(referred to as a D/A converter hereinafter) converts the corrected color/brightness signals from the field memory
16
into an analog image signal to be recorded.
In such an image correction system, covariance values are calculated between image data selected from two sequential images, and a motion vector is determined based on the point where the minimum covariance value occurs.
In general, a block matching algorithm is used to calculate the covariance value. Performance of the motion vector correction is impaired for the time it takes to apply the block matching algorithm. In order to completely calculate the correction under the constraints of real-time processing, it has been suggested that various techniques be used, such as pyramid searching, logarithmic searching, and so on. A two-dimensional block matching algorithm, however, entails a rapid increase in the number of arithmetic operations with an increase in the number of pixels to be considered. Furthermore, when a subset of pixels from the image is considered, the block matching algorithm is likely to produce erroneous results such as detecting a motion correction vector at a local minimum, which stabilizes only a portion of the image.
As shown in FIGS.
2
(A) and
2
(B), a general method has also been used which extracts one-dimensional components of the motion correction vector by projecting the image pattern onto a horizontal axis and a vertical axis, and computing the horizontal and vertical components of the correction vector separately.
After an image pattern has been projected in the directions of the horizontal and vertical axes, the projected result is compared with the projected result of the preceding image pattern to calculate one-dimensional covariance values. A displacement quantity between the two fields is determined from the point where the minimum covariance value results. Let an x and y coordinate system represent horizontal and vertical axes, respectively, of an image. Assuming that M is the number of horizontal pixels in each line of an image, and that N is the number of lines of an image, the horizontal covariance value, for example, can be calculated by an equation such as the following:
C
(
u
)
=
∑
x
=
S
M
-
S
&LeftBracketingBar;
P
h
′
(
x
+
u
)
-
P
h
(
x
)
&RightBracketingBar;
(
1
)
where P
h
′ and P
h
are line memories into which the preceding and the present image horizontal component data are accumulated, respectively. The variable u {u|−S<u<S} is an integer within a searching distance ±S, and it represents a displacement variable as a number of pixels. If a value of the variable u results in the minimum value of C(u), that value of u is considered to be the optimum horizontal displacement of an image. A similar computation is done for the vertical component. Compared with a two-dimensional matching algorithm, a one-dimensional signal matching algorithm applied to each of two axes x and y enables the calculation of a motion vector using fewer arithmetic operations, even when there is a large displacement between images.
The one-dimensional signal matching algorithm, however, also has a time restraint. The calculation of covariance should be finished before the raster scanning of the next field begins, yet the projection of the present field image data can not be completed until the raster scanning of a previous input image comes to an end. Further, an input pixel should be converted into a low quantified level (a binary signal, etc.) in order to economize on the memory used to accumulate the projection. Also, the determination of a threshold value in the conversion of pixels into a binary signal and the extraction of contours of an image, can lead to the loss of some pixel data. In addition, two pairs of line memories are required, of which one pair has N registers, the number of lines of an input image, and the other pair has M, the number of horizontal pixels. Finally, a complete one-dimensional signal matching algorithm must be implemented in the circuitry.
In order to solve the above-mentioned problems, Korean Patent Application No. 95-27157 has disclosed a device and method for detecting a motion vector of a camcorder. Besides a main CCD, the disclosed device uses separate image pick-up units to obtain one-dimensional component data at a high speed, uses a pipeline processing method to enable the rapid calculation of covariance values between successive image component data, and uses fewer line memories to allow for simpler circuitry.
FIG. 3
is a drawing which shows the principles of picking up an image through an image pick-up unit according to a conventional embodiment.
FIG. 4
is a drawing which shows horizontal and vertical image component pick-up units and peripheral units thereof.
As shown in
FIG. 3
, a conventional image pick-up unit
30
includes two right-angled prisms
31
and
32
, the hypotenuses of which are joined to each other to form a reflection surface that can reflect an incident two-dimensional image. A plano convex lens
33
has a flat surface which faces the reflection surface of the two right-angled prisms
31
and
32
, and condenses the reflected image into a one-dimensional image component. A line CCD
34
converts the one-dimensional image component into electrical signals.
The basi
Hwang Jung Hyun
Lee Chul Ho
Garber Wendy
Jones Volentine, LLC
White Mitchell
LandOfFree
Motion correction device for moving images does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Motion correction device for moving images, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Motion correction device for moving images will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2978793