Bad pixel detection and correction in an image sensing device

Details Bad pixel detection and correction in an image sensing device Bad pixel detection and correction in an image sensing device

2001-06-28

2004-05-18

Allen, Stephone B. (Department: 2878)

Radiant energy

Photocells; circuits and apparatus

Photocell controlled circuit

C348S230100, C348S615000

Reexamination Certificate

active

06737625

ABSTRACT:

BACKGROUND OF THE INVENTION
1. Technical Field of the Invention
The present invention relates to an image sensing device comprising an arrayed plurality of photodetectors and, in particular, to the detection of bad pixels in such a device and the correction of pixel data output from those bad pixels.
2. Description of Related Art
It is well known in the art to assemble a plurality of photodetectors (also referred to in the art as “photosites”) in an array format to create an image sensing device. Each individual photodetector operates to output a signal whose magnitude is proportional to the intensity of light incident on the site of the photodetector. These output signals can then be subsequently processed and manipulated to generate an image comprised of a plurality of individual picture elements (also referred to in the art as “pixels”), wherein each pixel in the image corresponds with one of the photodetectors.
The individual photodetectors used in such a sensing device are typically photodiodes that are formed on a semiconductor substrate or chip. The array of these photodetectors may include thousands (if not hundreds of thousands or millions) of individual photodetectors. Even with a semiconductor manufacturing process having a 99% yield, it is recognized that the sensing device array will inevitably include a number of bad photodetectors (also referred to herein as “bad pixels”). Requiring a semiconductor production line to produce sensors without any bad pixels is a very stringent and perhaps an economically unfeasible and unrealizable demand. It is accordingly accepted for sensors to include some bad pixels as long as a suitable mechanism is available to detect the bad pixels and satisfactorily correct the photodetector output signals. These detection/correction mechanisms generally comprise hardware or software processes that detect bad pixels based on their output signals and alter/replace their output signals to more accurately reflect the intensity of the light incident on the photodetector.
A bad pixel within a sensor may arise in a number of ways. First, the signal output from the photodetector may always give a high reading (a “hot pixel”). This defect produces a brighter than expected spot in the generated image. Second, the signal output from the photodetector may always give a low reading (a “dead pixel”). This defect produces a darker that expected spot in the generated image. Third, a signal proportional to the incident light may be generated, but (due, for example, to incorrect photodiode gain, dark current and/or offset) does not accurately represent of the incident light in the sense that it differs from signals generated by other similarly situated photodetectors (a “noisy pixel”). This defect produces an erroneous spot (color, intensity, contrast, and the like) in the generated image.
FIG. 1
shows a simplified schematic diagram of a sensor read-out operation in accordance with the prior art. A sensor array
10
includes a plurality of photodetectors
12
. Each photodetector
12
outputs a signal (referred to herein as “pixel data”) whose magnitude ideally has a predetermined monotonic relationship to the intensity of light incident on the photodetector. Although not explicitly shown, the generated signal for each pixel has a digital format (obtained through appropriate analog-to-digital conversion) normally ten to twelve bits in size.
A read-out buffer
14
is used to read out and capture either a horizontal row or vertical column of individual pixel data that is generated from the photodetectors. In this context, a horizontal row or a vertical column is referred to as a “line”
16
. Following capture of a line
16
of pixel data in the buffer
14
, each individual piece of pixel data
18
is processed by a bad pixel processor
20
. The processor
20
operates to serially examine each piece of pixel data
18
in a line
16
, detect instances where that piece of data has been output by a bad pixel, and in such cases effectuate a correction on the pixel data so that it more accurately represents the intensity of the light that is incident on the photodetector. Pixel data is then serially output
22
by the processor
20
for each pixel in either its original form
24
(i.e., as read out by the buffer
14
) or a modified form
24
′ (i.e., as corrected when a bad pixel is detected).
FIG. 2
shows a block diagram illustrating one known operation for bad pixel processing performed by processor
20
of FIG.
1
. The bad pixel processor
20
serially processes the pixel data
18
captured by the buffer
14
for one line
16
(see,
FIG. 1
) one pixel at a time (generically referred to as pixel “X”) in order to determine if each piece of the pixel data under examination is generated by a bad pixel. To assist in the making of this determination, the mechanism includes a data buffer
26
that stores the pixel data
18
(as captured in buffer
14
) for not only the pixel X under examination, but also for the pixels Y and Z that neighbor (e.g., are immediately adjacent to) the pixel X in the same read-out line
16
.
A detection algorithm
28
processes the pixel data
18
for pixels X, Y and Z and determines if the pixel data for pixel X differs from the pixel data for pixels Y and Z (either individually, by median or on average) by a deviation that exceeds a certain threshold. If not, the pixel data
18
for pixel X is assumed to represent valid pixel data generated from a good pixel, and is serially output
22
in its original form
24
. If there is an excessive deviation (beyond the threshold), the pixel data
18
for pixel X is assumed to be generated from a bad pixel, and a correction algorithm
30
is executed to modify the pixel data to produce modified pixel data
24
′ that attempts to more accurately represent the intensity of the light that is incident on the photodetector. As an example, the correction algorithm may replace the pixel data
18
for pixel X with pixel data
24
′ comprising the median or average value of the pixel data
18
for the adjacent pixels Y and Z on the same line
16
. As another example, the correction algorithm may replace the pixel data
18
for pixel X with pixel data
24
′ comprising the pixel data
18
for either one of the adjacent pixels Y or Z on the same line
16
. Numerous other operations for correcting pixel data from bad pixel are known in the art. Following correction, the modified data
24
′ for pixel X is passed to serial output
22
.
FIG. 3
shows a block diagram illustrating another known operation for bad pixel processing performed by the processor
20
of FIG.
1
. The bad pixel processor
20
serially processes the pixel data
18
captured by the buffer
14
for one line
16
(see,
FIG. 1
) one pixel at a time (generically referred to as pixel “X”) in order to determine if any of the pixel data under examination is generated by a bad pixel. To assist in the making of this determination, the processor
20
includes a data buffer
26
that stores the pixel data
18
(as currently captured in buffer
14
) for not only the pixel X under examination, but also for the pixels Y and Z that neighbor (e.g., are immediately adjacent to) the pixel X in the currently read-out line
16
(
n
). To farther assist in the making of this determination, the processor
20
includes a data buffer
32
storing the pixel data
18
that was previously captured in buffer
14
for an adjacent, previously read-out line
16
(
n−
1). More specifically, the pixel data
18
of interest comprises that pixel data for pixels A, B and C that most closely neighbor pixel X in the previously read-out line
16
(
n−
1).
A detection algorithm
28
processes the pixel data
18
for pixels A, B, C, X, Y and Z and determines if the pixel data for pixel X differs from the pixel data for pixels A, B, C, Y and Z (either individually, in pairs or groups, by median or on average) by a deviation that exceeds a certain threshold. If not, the pixel data
18
for pixel X is assumed to represent valid pixel data genera

Affiliated with

Also associated with

Rating

Bad pixel detection and correction in an image sensing device does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Bad pixel detection and correction in an image sensing device, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Bad pixel detection and correction in an image sensing device will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3212434