WO2006043563A1 - ノイズ除去装置 - Google Patents
ノイズ除去装置 Download PDFInfo
- Publication number
- WO2006043563A1 WO2006043563A1 PCT/JP2005/019138 JP2005019138W WO2006043563A1 WO 2006043563 A1 WO2006043563 A1 WO 2006043563A1 JP 2005019138 W JP2005019138 W JP 2005019138W WO 2006043563 A1 WO2006043563 A1 WO 2006043563A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image signal
- noise
- unit
- dark current
- component
- Prior art date
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N25/00—Circuitry of solid-state image sensors [SSIS]; Control thereof
- H04N25/60—Noise processing, e.g. detecting, correcting, reducing or removing noise
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N25/00—Circuitry of solid-state image sensors [SSIS]; Control thereof
- H04N25/60—Noise processing, e.g. detecting, correcting, reducing or removing noise
- H04N25/63—Noise processing, e.g. detecting, correcting, reducing or removing noise applied to dark current
Definitions
- the present invention relates to an apparatus for removing noise from an image sensor used for a digital camera, a mobile phone, or the like.
- Image sensors such as CCDs and CMOS sensors used in digital cameras and mobile phones have a noise pattern that is determined to some extent due to the influence of dark current flowing in the image sensor that constitutes the pixels even in the absence of light. (Dark current noise) occurs.
- Patent Document 1 discloses an apparatus for removing the dark current noise.
- the dark current noise component which is an image captured by the image sensor with the shutter closed, is compressed in advance by orthogonal transformation and quantization and stored in a memory, and the image is actually captured. Then, the dark current noise is removed by subtracting the dark current noise component decoded by the orthogonal inverse transform from the image signal.
- Orthogonal transformation and quantization which is a method for compressing dark current noise disclosed in Patent Document 1, uses a well-known method for image compression. This method uses the general property of images that information is biased to low-frequency components when image signals are orthogonally transformed. Even if high-frequency component information is cut by quantization, the image quality is not much. Does not have a big impact.
- Patent Document 1 Japanese Patent Laid-Open No. 11-298762
- the present invention has been made in order to solve the above-described problems.
- the dark current noise component is highly compressed while maintaining characteristics with many high frequency components.
- An object of the present invention is to provide a noise removing device that can be memorized.
- the dark current noise component varies greatly depending on the temperature and exposure time of the image sensor at the time of image capture and the gain multiplied when the image signal is extracted as an analog signal from the output of the image sensor, it was captured in advance.
- the magnitude of the dark current noise component and the magnitude of the dark current noise component included in the image component captured in the actual imaging state are different.
- the dark current noise removal device disclosed in Patent Document 1 includes compression and expansion processing, the dark current noise component is stored in advance and the dark signal noise component is subtracted as it is, so that the image signal force is also correct. There was a problem that dark current noise components could not be removed.
- the present invention has been made to solve the above-described problems, and an object of the present invention is to provide a noise removing device capable of accurately removing a dark current noise component contained in an image signal.
- the signal output from the image sensor includes various noises in addition to smear noise.
- fixed pattern noise has a high temperature dependence of the dark current that causes it, and when the temperature of the image sensor rises by 7 degrees, the magnitude of the solid pattern noise doubles.
- Fixed pattern noise increases in proportion to the exposure time. Therefore, since the fixed pattern noise of the signal that also outputs the image sensor force changes greatly depending on the temperature of the image sensor and the exposure time, the device described in Japanese Patent Application Laid-Open No. 2004-172925 is used when estimating the smear noise. There is a problem that the accuracy of removing smear noise is deteriorated due to the influence of the fixed pattern noise.
- the present invention has been made in order to solve the above-described problems, and noise that can remove each of a plurality of noise components included in a signal output from an image sensor with high accuracy.
- An object is to provide a removal device.
- An embodiment of the present invention in the first means relates to a noise removing device.
- This device obtains the distribution of the size of the first image signal of some or all of the pixels constituting the image sensor with respect to the first image signal captured by the image sensor in a state where light is blocked. Therefore, the distribution force also specifies a range including a peak of the magnitude taken by the first image signal, and a noise distribution analysis unit that sets a threshold value within the range, and the imaging element in a state where light is blocked.
- a quantization unit that quantizes the second image signal imaged by the first threshold based on the threshold, an inverse quantization unit that inversely quantizes the quantized second image signal, and incident light And a subtracting unit that subtracts the inversely quantized second image signal from the third image signal imaged by the imaging element in this state.
- the “light-blocked state” refers to a state in which light is not transmitted to the image sensor by closing the shutter provided on the front surface of the image sensor. Includes a “substantially blocked light” state where the light is incident on the image sensor.
- noise component Most of the first and second image signals imaged in a state where light is blocked are noise components due to dark current of the imaging element. Therefore, according to this aspect, this noise component
- the amount of information on the noise component can be reduced, the memory capacity required to store the noise component can be reduced, and the size of the noise component when quantizing can be reduced.
- the distribution power of the noise component is determined, the range where the size of the noise component is concentrated is identified, and quantization compression is performed within that range, so the noise component can be highly compressed while maintaining the characteristics of noise with many high frequency components. It is possible to memorize it.
- the first image signal and the second image signal may be taken at different times. According to this, since the quantization threshold value is set and quantized using two image signals captured at different times, it is not necessary to temporarily store the captured image signal. Memory capacity can be reduced by the amount.
- the first image signal and the second image signal are the same image signal
- the noise distribution analyzer divides the first image signal into a plurality of regions, The distribution of the magnitude of the first image signal is obtained for each divided area, the range including the peak of the magnitude taken by the first image signal is identified from this distribution, and the threshold value is set within this range. May be. According to this, since the noise component, which is the captured image signal, is stored, the number of times that the image signal is captured in a state where light is blocked is only one, so that the quantization of the dark current noise component is simple and short. It is possible to do in time.
- the apparatus further comprises a band separation unit that divides the first image signal and the second image signal into a low-frequency component and a high-frequency component, and the noise distribution analysis unit for the high-frequency component And adding the quantization unit and the inverse quantization unit, adding the low frequency component and the high frequency component inversely quantized by the inverse quantization unit, and inputting this to the subtraction unit May be.
- the noise component differs from pixel to pixel, the high-frequency component is dominant, but depending on the device characteristics and power supply, the low-frequency component may be mixed into the acquired noise information. In such a case, it is possible to restore the noise component more accurately by separating the dark current noise into a low frequency component and a high frequency component and performing the compression processing of the present invention on this high frequency component. is there.
- the apparatus further comprises a compression unit that compresses the low-frequency component by a method independent of the high-frequency component, and an expansion unit that expands the compressed low-frequency component. After adding the component and the high frequency component dequantized by the dequantization unit, May be input to the subtracting unit. According to this, by compressing the low-frequency component and the high-frequency component by independent methods, it is possible to restore the noise component more accurately because the compression can be performed by a method suitable for the characteristics of each component. .
- Another aspect of the present invention in the second means relates to a noise removal method.
- the step of obtaining the distribution of the magnitude of the dark current noise component with respect to the image signal captured by the imaging device in a state where light is blocked, and the range including the peak of the magnitude of the dark current noise component are specified.
- the amount of information of the dark current noise component can be reduced by quantizing the dark current noise component, and the memory capacity required to store the dark current noise component can be reduced. can do.
- the distribution of the magnitude of the dark current noise component is obtained, the range where the magnitude of the dark current noise component is concentrated is identified from the distribution, and quantization compression is performed within that range.
- An aspect of the present invention in the second means relates to a noise removal apparatus.
- the apparatus includes: a compression unit that compresses the first image signal obtained by the imaging element in a state where light is blocked; a decompression unit that decompresses the compressed first image signal; and a compression unit that compresses the compressed first image signal. Based on a comparison between the state of the first image signal before being imaged and the state of the second image signal obtained by the imaging device in the state where light is incident, the first image signal is expanded from the expanded first image signal.
- a noise component prediction unit that predicts a noise component included in the second image signal; and a subtraction unit that subtracts the predicted noise component from the second image signal.
- the state in which light is blocked refers to a state in which light does not reach the image sensor by closing a shutter provided in front of the image sensor, and is not completely blocked. It also includes “a state where light is substantially blocked” such that light leaked into the image sensor is incident on the image sensor. In addition, the first image signal is almost equal to the noise component caused by the dark current when this is imaged.
- the image signal obtained according to the condition for capturing the first image signal Since the relationship between the magnitude and the magnitude of the image signal obtained according to the conditions under which the second image signal is captured can be determined, the first image signal is changed to the second image signal based on this relation.
- the included noise component can be predicted. Therefore, the second image force can also accurately remove the noise component by subtracting the predicted noise component.
- the state of the first image signal is a magnitude of a signal output from a light shielding region force included in the first image signal before being compressed
- the second image The state of the signal may be the magnitude of the signal output from the light shielding area force included in the second image signal.
- the noise component prediction unit outputs the light shielding area force included in the second image signal and the magnitude of the output signal and the light shielding area included in the first image signal before being compressed.
- a noise component included in the second image signal may be predicted by obtaining a ratio to the magnitude of the signal and multiplying the expanded first image signal by the ratio.
- a signal output from a region in which light included in the second image signal is not incident is substantially equal to a noise component caused by the dark current generated under the condition of capturing the second image signal.
- the signal output from the region where the light included in the first image signal is not incident is substantially equal to the noise component caused by the dark current generated under the condition where the first image signal is captured. Therefore, the noise component contained in the second image signal can be predicted by calculating the ratio with the magnitude of these signals and multiplying this ratio by the first image signal. By subtracting the predicted noise component from the image source, it is possible to remove the noise component with high accuracy.
- the image processing apparatus further includes a preprocessing unit that performs preprocessing for predicting a noise component included in the second image signal with respect to the compressed first image signal, and the extension.
- the unit may expand the preprocessed first image signal. According to this, the pre-processing for predicting the noise component included in the second image signal is performed on the first image once compressed.
- the image signal is not compressed but is compressed. Since the compressed image signal has a smaller data amount than the expanded image signal, the amount of calculation can be reduced.
- the pre-processing unit separates the compressed first image signal into a low-frequency component and a high-frequency component, and the expansion unit includes the low-frequency component.
- the noise component prediction unit predicts the high frequency component of the noise component included in the second image signal from the expanded high frequency component and then expands the low frequency component.
- the noise component included in the second image signal may be predicted by adding the component.
- dark current noise included in an image signal contains many high-frequency components, and is greatly influenced by imaging conditions such as temperature and exposure time.
- low frequency components are dominated by noise due to power supply and device characteristics, and this is relatively less affected by imaging conditions such as temperature and exposure time.
- the prediction process is performed on the high-frequency component having a large influence of the imaging condition except for the low-frequency component having a small influence of the imaging condition, a more accurate noise component can be predicted. Therefore, noise components included in the image can be accurately removed.
- the state of the first image signal is the magnitude of the high-frequency component of the signal output from the light shielding region force included in the first image signal before compression.
- the state of the second image signal is the magnitude of the high-frequency component of the signal output from the light shielding region included in the second image signal, and the noise component prediction unit is included in the second image signal.
- the ratio of the high frequency component of the outputted signal to the high frequency component of the output signal and the size of the high frequency component of the output signal of the light shielding region force output included in the first image signal before being compressed is obtained.
- the high frequency component of the noise component contained in the second image signal may be predicted by multiplying the high frequency component of the first image signal by the ratio.
- the compression unit obtains a distribution of the size of the first image signal of some or all of the pixels with respect to the first image signal, and distributes the distribution power of the first image signal.
- a range including a peak of the magnitude taken by the image signal is specified, and a noise distribution analysis unit for setting a quantization threshold within the range, and a quantizer based on the threshold for the first image signal. And a quantization unit that performs the conversion. According to this, when storing the first image signal, quantization compression is performed within a range including the peak of the size taken by the first image signal, so that the noise characteristic that there are many high frequency components is maintained.
- the first image signal can be compressed and stored.
- An aspect of the present invention according to the third means relates to a noise removing device.
- This apparatus includes a fixed pattern noise removing unit that removes fixed pattern noise caused by dark current of the image sensor from the image signal output from the image sensor, and an incident light from the image signal from which the fixed pattern noise has been removed.
- a smear noise removing unit that removes smear noise caused by the smear noise.
- the fixed pattern noise force that has temperature dependence and increases in proportion to the exposure time is removed from the image signal before removing the smear noise. Therefore, the smear noise can be estimated using the image signal from which the fixed pattern noise has been removed, and the smear noise can be accurately removed.
- the noise removing device described in the first means or the second means described above may be used. According to this aspect, it is possible to effectively remove fixed pattern noise.
- the smear noise removing unit receives each light reception from the image signal from which the fixed pattern noise has been removed, every time information charges accumulated in each light receiving bit of the image sensor are transferred in a vertical direction. Smear noise may be removed by subtracting a value obtained by sequentially adding the amount of smear charge mixed from the bit. According to this aspect, it is possible to effectively remove smear noise.
- a fixed defect noise removing unit that removes fixed defect noise caused by defects at the time of manufacturing the image sensor from the image signal from which the smear noise has been removed may be further provided. Yes.
- the smear noise is removed before the fixed defect noise is removed, it is possible to determine a defective pixel that is not affected by the signal level saturated by the smear noise.
- the interpolated image signal can be generated from the image signal from which fixed pattern noise with a large noise level has been removed, so that fixed defect noise can be removed more naturally. It becomes.
- the fixed defect noise removing unit determines whether or not each pixel of the image sensor is a fixed defect, and if it is determined as a fixed defect, calculates a peripheral pixel force interpolation value and determines the fixed defect. It may be replaced with a value. According to this aspect, fixed defect noise can be effectively removed.
- the image processing apparatus may further include a random noise removing unit that removes random noise caused by thermal fluctuation of the image sensor from the image signal from which the fixed defect noise has been removed.
- the image processing apparatus further includes an offset removal unit that removes an offset component included in the image signal, and the image signal input to the offset removal unit is obtained by removing fixed pattern noise by the fixed pattern noise removal unit. There may be.
- the offset component since the offset component is calculated after removing the fixed pattern noise, the offset component can be accurately removed without being affected by the fixed pattern noise in the calculation of the offset component. Become.
- the offset removal unit extracts an image signal of a pixel belonging to the light-shielding region of the imaging element from the image signal input to the offset removal unit, and calculates an average value of the size of the extracted image signal.
- the noise removal performance can be improved.
- FIG. 1 is a configuration diagram of a digital camera 100 according to a first embodiment.
- FIG. 2 is a diagram showing the distribution of the magnitude of dark current noise components.
- FIG. 3 is a diagram for explaining that a range including a peak of a magnitude taken by a dark current noise component is specified, and a quantization threshold and a representative value are calculated within the range.
- FIG. 4 is a configuration diagram of a digital camera 110 according to a second embodiment.
- FIG. 5 is a diagram for explaining that a dark current noise component is divided into a plurality of lines.
- FIG. 6 is a configuration diagram of a digital camera 120 according to a third embodiment.
- FIG. 7 is a configuration diagram of a digital camera 1100 according to a fourth embodiment.
- FIG. 8 is a diagram showing the distribution of the magnitude of dark current noise components.
- FIG. 9 is a diagram for explaining that a range including a peak of a magnitude taken by a dark current noise component is specified, and that a quantization threshold and a representative value are calculated within the range.
- FIG. 10 is a configuration diagram of a digital camera 2100 according to a fifth embodiment.
- FIG. 11 is a block diagram of a fixed pattern noise removing unit 2010 according to the fifth embodiment.
- FIG. 12 is a configuration diagram of a smear noise removing unit 2011 according to the fifth embodiment.
- FIG. 13 is a block diagram of a fixed defect noise removing unit 2012 according to Embodiment 5.
- FIG. 14 is a diagram for explaining a fixed defect determination method.
- FIG. 15 is a configuration diagram of a digital camera 2110 according to the sixth embodiment.
- FIG. 16 is a configuration diagram of an offset removal unit 2014 according to Embodiment 6.
- FIG. 1 is a diagram showing a configuration of a digital camera 100 provided with a dark current noise removing apparatus 1 according to a preferred embodiment 1 of the present invention.
- This configuration can be realized in hardware by a CPU, memory, or other LSI of any computer, and in software, it can be realized by a program with an encoding function loaded in the memory. It draws functional blocks that are realized by their cooperation. Therefore, those skilled in the art will understand that these functional blocks can be realized in various forms by hardware only, software only, or their combination.
- the digital camera 100 includes a dark current noise removing device 1, a lens 2, a shutter 3, an imaging element 4, an AZD conversion unit 5, and a recording medium 6.
- the light reflected by the subject enters the image sensor 4 through the lens 2 and the shutter 3.
- the image sensor 4 converts the incident light into an electrical signal and outputs it as an image signal.
- Examples of the image pickup device 4 include a CCD and a CMOS sensor.
- the image signal output from the image sensor 4 is converted into, for example, a 10-bit digital signal by the AZD conversion unit 5, and then the dark current noise is removed by the dark current noise removal device 1 and recorded on the recording medium 6. Is done.
- the dark current noise removing device 1 uses an image signal that has been picked up by the image pickup device 4 with the shutter 3 closed, and converted into a digital signal by the AZD conversion unit 5 as a current noise. It is stored as a component, and this power is calculated from the image signal captured during actual imaging. The dark current noise is removed by subtracting the current noise component. When storing this dark current noise component, the distribution of the magnitude of the dark current noise component of all pixels constituting the image is obtained, and the range including the peak of the dark current noise component is examined from this distribution, and the quantum within that range is determined. Perform compression.
- the magnitude of the dark current noise component referred to here is a value represented by a digital signal obtained by the AZD conversion unit 5.
- the dark current noise removal apparatus 1 includes a noise distribution analysis unit 10, a quantization unit 12, a memory 14, an inverse quantization unit 16, and a subtraction unit 18.
- the noise distribution analysis unit 10 counts for each pixel how many pixels appear for each possible value in all pixels, Find its distribution.
- the dark current noise component of an image composed of 10 pixels has the following values for each pixel.
- the noise distribution analysis unit 10 obtains the number of appearing pixels of each value as follows.
- Value 5 2 pixels
- value 6 0 pixels
- value 7 0 pixels.
- the number of pixels of the image actually obtained from the image sensor 4 is several hundred thousand or even millions of pixels.
- the value that the dark current noise component can take is determined by the resolution of the AZD converter 5. For example, if the AZD converter 5 has a 10-bit resolution, what value can the dark current noise component take? ⁇ 1023.
- the noise distribution analysis unit 10 obtains the number of appearing pixels for each possible value of 0 to 1023 for the dark current noise components of all the pixels obtained by the image sensor 4. As a result, the distribution shown in Fig. 2 can be obtained.
- the noise distribution analysis unit 10 identifies ranges A to B where the dark current noise components are concentrated as shown in FIG.
- the lower limit A of this range is the value of the dark current noise component having the pixel power of the xZ512 pixel, counting from the smaller of the pixel values, where X is the number of pixels used to determine the distribution.
- the upper limit B is the value of the dark current noise component of the xZ512 pixel, counting from the largest pixel value.
- the divisor is not limited to 512, and may be an arbitrary value. Different divisors may be used for the lower limit A and the upper limit B. Furthermore, the divisor may be variable and set from the outside.
- the position of the peak of the distribution of the dark current noise component is specified, the number of pixels existing from the lower limit A to this peak, and the upper limit B Force Number of pixels existing up to this peak
- the lower limit A and the upper limit B may be set so that a certain number of pixels is obtained.
- THn DA + (2n + l) ⁇ (DB-DA) / (2- (N—l)) ⁇ ⁇ ⁇ ⁇ (1)
- Vn DA + 2n- (DB-DA) / (2- (N-l))
- DA and DB represent the values of the dark current noise component at the lower limit A and the upper limit B, respectively.
- the threshold value THn is sent to the quantization unit 12, while the representative value Vn is sent to the inverse quantization unit 16.
- the quantization unit 12 quantizes the dark current noise component obtained when the shutter 3 is closed and refers to the quantization threshold obtained by the noise distribution analysis unit 10. .
- quantization is performed by the method described below while referring to the threshold value THn obtained by Equation (1) to obtain post-quantization data OC.
- the quantized data ⁇ is set to zero. If the dark current noise component a is greater than or equal to the threshold ⁇ ( ⁇ 2), the quantized data ⁇ is (N ⁇ l). The dark current noise component a is greater than or equal to the threshold TH (n-l) and less than THn. In this case, the quantized data is n.
- the noise distribution analysis unit 10 and the quantization unit 12 are examples of the “means for performing quantization of noise components” in the present invention.
- the data ⁇ quantized by the quantization unit 12 is stored in the memory 14.
- the capacity required for the memory 14 may be smaller than that required when the dark current noise component is quantized.
- the quantized data a takes a value from 0 to 15. That is, the dark current noise component that was originally represented by 10 bits per pixel can be represented by 4 bits after quantization. Therefore, if a dark current noise component of 1 million pixels is stored in the memory 14 if it is not quantized, a capacity of 10 million bits is required, but if it is quantized, 4 million bits are enough.
- the inverse quantization unit 16 is a dark current quantized from the memory 14 in accordance with the timing at which an image signal captured with the shutter 3 opened during actual imaging is output from the AZD conversion unit 5.
- the noise component is read out and dequantized and decoded while referring to the representative value input from the noise distribution analyzer 10.
- the subtracting unit 18 subtracts the dark current noise component inversely quantized by the inverse quantizing unit 16 from the image signal force captured at the time of actual imaging.
- the image sensor 4 captures an image with the shirt 3 closed. At this time, light is incident on the image sensor 4.
- the signal output from the image sensor 4 is a dark current noise component.
- This dark current noise component is converted into a digital signal by the AZD conversion unit 5 and then input to the noise distribution analysis unit 10.
- the noise distribution analysis unit 10 obtains the distribution of the dark current noise component force and magnitude of all pixels, examines the range including the peak of the dark current noise component from this distribution, and represents the quantization threshold THn within that range as a representative. Calculate the value Vn.
- the dark current noise component is imaged by the image sensor 4 with the shutter 3 closed again.
- This dark current noise component is converted into a digital signal by the AZD converter 5 and then input to the quantizer 12.
- the quantization unit 12 is obtained by the noise distribution analysis unit 10.
- the dark current noise component is quantized based on the threshold value, and the memory 14 stores the quantized dark current noise component.
- the above operation is the operation up to the stage before actually imaging the subject.
- the operation so far, that is, storing the quantized dark current noise component in the memory 14 may be performed at every imaging, or may be performed when the digital camera is turned on.
- it may be configured to provide an imaging time detection unit (not shown) and instruct the current noise removal device 1 to quantize and store the dark current noise component when the imaging time is detected.
- dark current noise is easily affected by the temperature environment around the image sensor, the dark current noise component value can be obtained with the highest accuracy when stored immediately before imaging.
- a power-on detection unit (not shown) is provided, and if power-on is detected, the dark current noise removal device 1 is instructed to quantize and store the dark current noise component. Good.
- the quantized dark current noise component may be stored in the memory 14 when the digital camera is manufactured.
- a timer (not shown) may be provided in the digital camera so that the dark current noise removal device 1 is instructed to quantize and store the dark current noise component at regular intervals.
- the image sensor 4 captures an image with the shutter 3 opened.
- the image signal output from the image sensor 4 is converted into a digital signal by the AZD converter 5 and then sent to the subtractor 18.
- the inverse quantization unit 16 reads the quantized dark current noise component stored in the memory 14 in accordance with the timing at which the image signal is output from the AZD conversion unit 5, and the noise distribution analysis unit 10
- the dark current noise component is inversely quantized while referring to the calculated representative value.
- the inversely quantized dark current noise component is input to the subtracting unit 18, and the subtracting unit 18 subtracts the inversely quantized dark current noise component from the image signal power.
- the negative current noise component can be removed from the image signal.
- the image signal from which the dark current noise component has been removed is recorded on the recording medium 6.
- the amount of information of the dark current noise component is greatly increased by quantizing the dark current noise component.
- the amount of memory required to store the dark current noise component can be reduced.
- the quantization threshold value is set and quantized using two dark current noise components captured at different times, the captured dark current noise components are temporarily stored. Memory capacity that is not necessary can be reduced.
- the quantization threshold is set using the dark current noise component obtained first, but it is only necessary to retain the dark current noise component for the moment of the quantization threshold setting. You may delete the data.
- quantization is performed using the dark current noise component acquired next, but it is only necessary to retain the dark current noise component for the moment of the quantization, and thereafter erase the dark current noise component data. Also good. Since it is not necessary to maintain a dark current noise component between the quantization threshold setting process and the quantization process, the memory capacity can be reduced.
- FIG. 4 is a diagram showing a configuration of a digital camera 110 provided with the dark current noise removing apparatus 7 according to the preferred embodiment 2 of the present invention.
- the digital camera 110 has a configuration in which a noise distribution analysis unit 20 having a function different from that of the noise distribution analysis unit 10 of the digital camera 100 according to the first embodiment is arranged, and a line memory 22 is further added.
- the same components as those in the first embodiment are denoted by the same reference numerals and description thereof is omitted.
- the noise distribution analysis unit 20 in the second embodiment is different from the noise distribution analysis unit 10 in the first embodiment in that the distribution of the magnitude of the dark current noise component is performed over all pixels. Then, the dark current noise component is divided into multiple line areas as shown in Fig. 5, the distribution of the magnitude of the dark current noise component is obtained within the divided range, and the quantization threshold and representative value are calculated. It is in the point to put out. In addition, the noise distribution analysis unit 20 stores the dark current noise components of a plurality of lines for which the quantization threshold and the representative value are calculated in the line memory 22.
- the image sensor 4 images the dark current noise component.
- This dark current noise component is converted into a digital signal by the AZD converter 5 and then input to the noise distribution analyzer 20.
- the noise distribution analysis unit 20 obtains the magnitude distribution from the dark current noise components of the pixels in the first plurality of lines, and calculates the quantization threshold and the representative value based on this distribution. Further, the noise distribution analysis unit 20 obtains the distribution, and stores the dark current noise components of a plurality of lines whose quantization thresholds and representative values are calculated in the line memory 22.
- the quantization unit 12 reads the dark current noise component stored in the line memory 22, performs quantization while referring to the threshold value calculated by the noise distribution analysis unit 20, and performs this quantization.
- the dark current noise component is stored in the memory 14.
- the noise distribution analysis unit 20 obtains the distribution of the magnitude of the dark current noise component for the pixels existing in the next multiple lines, and calculates the quantization threshold and the representative value from this distribution.
- the dark current noise components of the plurality of lines are overwritten in the line memory 22.
- the noise distribution analysis unit 20, the quantization unit 12, and the line memory 22 repeat the above operation for each of a plurality of lines to perform quantization compression on all the pixels of the dark current noise component.
- the quantized dark current noise component can be stored in the memory 14.
- the above operation is the operation up to the stage before actually imaging the subject. Since the operation when actually imaging the subject is the same as the operation described in the first embodiment, the description is omitted.
- the dark current noise component is divided into a plurality of regions, the distribution of the magnitude of the dark current noise component is obtained for each region, the quantization threshold value and the representative value are calculated, and Line memo is the dark current noise component of the area where the representative value is calculated. Since the dark current noise component stored in the line memory is quantized, the dark current noise component can be imaged only once, and the dark current noise component can be quantized. It can be performed easily and in a short time.
- FIG. 6 is a diagram showing a configuration of a digital camera 120 including the dark current noise removing device 8 according to the preferred embodiment 3 of the present invention.
- the digital camera 120 has a configuration in which a band separation unit 24, a compression unit 28, a memory 30, an expansion unit 32, and an addition unit 34 are added to the digital camera 100 according to the first embodiment.
- the same components as those in the first embodiment are denoted by the same reference numerals and description thereof is omitted.
- the band separation unit 24 separates the dark current noise component into a low frequency component and a high frequency component.
- the compression unit 28 performs compression processing on the low-frequency component separated by the band separation unit 24.
- compression methods there are a method of taking a difference between low frequency components between a pixel to be compressed and an adjacent pixel, and a method of dividing and quantizing a low frequency component by a predetermined quantization coefficient. Alternatively, a combination of these methods may be used in which low-frequency components are extracted at predetermined pixel intervals and low-frequency components of other pixels are discarded.
- the memory 30 stores the low frequency component compressed by the compression unit 28.
- the expansion unit 32 reads the compressed low frequency component stored in the memory 30 and performs expansion processing. For example, when the low frequency component is compressed by the compression unit 28 by the method of taking the difference between the adjacent pixel to be compressed and the low frequency component, the value stored in the memory 30 and the low frequency component of the adjacent pixel are calculated. The expansion process is performed by adding. In addition, when compressed by a method of quantization by dividing by a predetermined quantization coefficient, expansion processing is performed by multiplying a value stored in the memory 30 by a quantization coefficient. In addition, when the low frequency components are extracted at a predetermined pixel interval and compressed by a method of discarding the low frequency components of other pixels, the low frequency components of the pixels stored in the memory 30 are filtered. Thus, the low frequency components of all pixels are obtained and the expansion process is performed.
- the noise distribution analysis unit 10, the quantization unit 12, the memory 14, and the inverse quantization unit 16 have the same functions as those provided in the digital camera according to the first embodiment. It works against the high-frequency component of the dark current noise separated at the separation part 24. [0082]
- the adder 34 adds the low-frequency component of the dark current noise expanded by the expander 32 and the high-frequency component of the dark current noise inversely quantized by the inverse quantization unit 16 to add dark current noise. Is decrypted.
- the dark current noise component imaged with the shirt 3 closed is converted into a digital signal by the AZD conversion unit 5 and then separated into a low frequency component and a high frequency component by the band separation unit 24.
- the noise distribution analysis unit 10 obtains a distribution of the size of the high frequency component over all pixels, and calculates a quantization threshold and a representative value for the high frequency component based on this distribution.
- the dark current noise component is imaged by the image sensor 4 with the shutter 3 closed again, converted into a digital signal by the AZD converter 5, and then the low frequency component by the band separator 24. And high frequency components.
- the low frequency component is compressed by the compression unit 28, and the compressed low frequency component is stored in the memory 30.
- the high frequency component is input to the quantization unit 12, quantized based on the threshold value obtained by the noise distribution analysis unit 10, and the quantized high frequency component is stored in the memory 14.
- the above operations are operations up to the stage before actually imaging the subject. Next, the operation when actually imaging a subject will be described.
- the image sensor 4 captures an image with the shutter 3 opened.
- the image signal output from the image sensor 4 is converted into a digital signal by the AZD conversion unit 5 and then sent to the subtraction unit 18.
- the decompression unit 32 reads out the low-frequency component of the quantized dark current noise stored in the memory 30 in accordance with the timing at which the image signal is output from the AZD conversion unit 5, and the dark current noise. Stretch low frequency components.
- the inverse quantization unit 16 reads the high-frequency component of the quantized dark current noise stored in the memory 14 in accordance with the timing at which the image signal is output from the AZD conversion unit 5.
- the high frequency component of the dark current noise is inversely quantized while referring to the representative value calculated by the noise distribution analysis unit 10.
- the dark current noise component is decoded by adding the inversely quantized low frequency component and high frequency component by the adding unit 34.
- the dark current noise component is input to the subtracting unit 18, and the dark current noise component obtained by dequantizing the image signal force by the subtracting unit 18 is subtracted to remove the dark current noise component from the image signal. Leave.
- the dark current noise itself varies from pixel to pixel, the high frequency component is dominant.
- the low frequency component may be mixed into the acquired noise information. According to the third embodiment, the following operational effects can be enjoyed.
- the dark current noise is separated into a low frequency component and a high frequency component, the distribution of the magnitude is obtained for the dominant high frequency component in the dark current noise, and the dark current noise is calculated from the distribution. It is possible to restore dark current noise more accurately by specifying a range that includes the peak of the magnitude taken by the high-frequency component and performing quantization compression within that range.
- the force shown in the example of the digital camera is not limited thereto, and the dark current noise removing device according to the embodiment of the present invention can be provided as long as the image sensor is provided. .
- the quantization threshold is determined in the divided range, and the quantization is performed is shown.
- the quantization threshold is determined from one dark current noise component of two dark current noise components captured at different times, and the other dark current noise is determined. It may be possible to select which quantization method to use, which may add a process for quantizing the component.
- the line memory is divided into noise distribution components as in the second embodiment.
- a low frequency component and a high frequency component of a plurality of lines used in the noise distribution analysis units 10 and 26 may be stored in these line memories by being added between the analysis unit 10 and the quantization unit 12.
- the dark current noise component can be divided into a plurality of lines, and the quantization threshold of the dark current noise component can be set and quantized within the divided range. It is possible to enjoy the same effects as those of the second embodiment.
- FIG. 7 is a diagram showing a configuration of a digital camera 1100 provided with a dark current noise removing apparatus 1001 according to a preferred embodiment 4 of the present invention.
- This configuration can be realized in hardware by any computer's CPU, memory, or other LSI, and in software, it can be realized by a program with an encoding function loaded in memory. Describes functional blocks realized by collaboration. Therefore, those skilled in the art will understand that these functional blocks can be realized in various forms by hardware only, software only, or a combination thereof.
- the digital camera 1100 includes the power of the eddy current noise removing device 1001, a lens 1002, a shirter 1003, an image sensor 1004, an AZD conversion unit 1005, and a recording medium 1006.
- the light reflected by the subject enters the image sensor 1004 through the lens 1002 and the shutter 1003.
- the image sensor 1004 converts the incident light into an electric signal and outputs it as an image signal. Examples of the image sensor 1004 include a CCD and a CMOS sensor.
- the image signal output from the image sensor 1004 is converted into, for example, a 10-bit digital signal by the AZD conversion unit 1005.
- the value represented by this digital signal corresponds to “the magnitude of the image signal” of the present invention. Further, when an image signal converted into a digital signal is separated into a low frequency component and a high frequency component, the value of each component represented by the digital signal is also included in the “image signal size”.
- the image signal converted into a digital signal by the AZD conversion unit is recorded on the recording medium 1006 after dark current noise is removed by the dark current noise removing device 1001.
- the dark current noise removing apparatus 1001 is imaged by the imaging element 1004 with the shutter 1003 closed, and converted into a digital signal by the A / D conversion unit 1005.
- the image signal is stored as a dark current noise component, and the image signal force captured at the time of actual imaging is subtracted from the dark current noise component to remove the dark current noise.
- the distribution of the magnitude of the dark current noise component of all pixels constituting the image is obtained, and the range including the peak of the dark current noise component is examined from this distribution, and the quantum within that range is determined. Perform compression.
- the magnitude of the dark current noise component mentioned here is a value represented by a digital signal obtained by the AZD conversion unit 1005.
- the dark current noise removing apparatus 1001 compares the state when the dark current noise component is captured in advance with the state when the image signal is captured. Based on this comparison, the dark current noise component included in the image signal is predicted from the dark current noise component captured in advance. For example, since there is a region where light is not always incident in the image sensor 1004, an image signal output from the region can be regarded as a dark current noise component in the pixel. Therefore, among the image signals captured in the light incident state, the image signal output from the pixel belonging to the region where the light is not always incident (the light shielding region) and the dark current noise component captured in advance from the same pixel. By calculating the ratio of the output dark current noise component and multiplying this ratio by the entire dark current noise component that has been imaged in advance, the dark current noise component included in the image signal is accurately predicted.
- the dark current noise removing apparatus 1001 separates the dark current noise and the image signal output from the light-shielding area, which are captured in advance, into a low frequency component and a high frequency component, and compares these high frequency components with each other.
- the above-described prediction is performed using dark current high-frequency components imaged in advance. This is because dark current noise varies from pixel to pixel and contains a lot of high-frequency components and is not affected by location. On the other hand, low-frequency components are affected by location such as power supply and device characteristics. This is because the noise is dominant, so if the prediction process including the low frequency component is performed, the dark current noise cannot be predicted correctly.
- the dark current noise removal apparatus 1011 includes a noise distribution analysis unit 1010, a quantization unit 1012, a memory 1014, a first inverse quantization unit 1018, a second inverse quantization unit 1020, a high frequency extraction unit 1022, a multiplication coefficient setting unit 1024.
- Noise distribution analyzer 1010 is a shirt For the dark current noise component obtained when the image is taken with the shutter 1003 closed, the number of pixels that appear for each possible value is counted for all pixels, and the distribution is obtained.
- the dark current noise component of an image composed of 10 pixels has the following values for each pixel.
- the noise distribution analysis unit 1010 obtains the number of appearing pixels of each value as follows.
- Value 5 2 pixels
- value 6 0 pixels
- value 7 0 pixels.
- the number of pixels of an image actually obtained from the image sensor 1004 is several hundred thousand or even millions of pixels.
- the value that the dark current noise component can take is determined by the resolution of the AZD converter 1005.
- the values that the dark current noise component can take are 0 to 1023.
- the noise distribution analysis unit 1010 obtains the number of appearing pixels of the possible values 0 to 1023 for the dark current noise components of all the pixels obtained by the imaging element 1004. As a result, the distribution shown in Fig. 8 can be obtained.
- the noise distribution analysis unit 1010 identifies ranges A to B including the peak of the dark current noise component as shown in FIG.
- the lower limit A of this range is the value of the dark current noise component having the pixel power of the xZ512 pixel, counting from the smaller of the pixel values, where X is the number of pixels used to determine the distribution.
- the upper limit B is the value of the dark current noise component of the xZ512 pixel, counting from the largest pixel value.
- the divisor is not limited to 512, and may be an arbitrary value. Different divisors for the lower limit A and the upper limit B may be used. Furthermore, the divisor may be variable and set from the outside.
- the position of the peak of the distribution of the dark current noise component is specified, the number of pixels existing from the lower limit A to this peak, and the upper limit B. Force Number of pixels existing up to this peak
- the lower limit A and the upper limit B may be set so that a certain number of pixels is obtained.
- the noise distribution analysis unit 1010 specifies the ranges A to B, and calculates a quantization threshold and a representative value so as to quantize and compress the threshold within the range. For example, consider the case where the dark current noise component is quantized into N stages as shown in Fig. 9.
- quantization level n 0, 1, 2,..., N ⁇ 1
- the noise distribution analysis unit 1010 obtains the threshold value THn between the quantization level n and the quantization level (n + 1) and the representative value Vn of the quantization level n by the following equation.
- THn DA + (2n + l) ⁇ (DB-DA) / (2-(N— 1)) ⁇ ⁇ ⁇ ⁇ (3)
- Vn DA + 2n- (DB-DA) / (2-(N-l)) ⁇ ⁇ ⁇ (4)
- DA and DB represent the values of the dark current noise component at the lower limit A and the upper limit B, respectively.
- the threshold value T Hn is sent to the quantization unit 1012.
- the dark current noise component DA at the threshold THn, the representative value Vn, and the lower limit A is sent to the first inverse quantization unit 1018 and the second inverse quantization unit 1020.
- the quantization unit 1012 quantizes the dark current noise component obtained when the shutter 1003 is closed and refers to the quantization threshold obtained by the noise distribution analysis unit 1010. .
- quantization is performed by the method described below while referring to the threshold value THn obtained by Equation (3) to obtain post-quantization data a.
- the quantized data a is set to zero. If the dark current noise component a is greater than or equal to the threshold TH (N ⁇ 2), the quantized data ⁇ is (N ⁇ l). If the dark current noise component a is greater than or equal to the threshold TH (n ⁇ l) and less than THn, the quantized data ex is n.
- the data ⁇ quantized by the quantization unit 1012 is stored in the memory 1014.
- the capacity required for the memory 1014 may be smaller than that required when the dark current noise component is quantized.
- the quantized data ⁇ takes a value from 0 to 15.
- the dark current noise component that was originally represented by 10 bits per pixel can be represented by 4 bits after quantization. Therefore, if a dark current noise component of 1 million pixels is stored in the memory 1014, if it is not quantized, 10 It needs 4 million bits when it is quantized while it needs a capacity of 400 million bits
- the low-frequency / high-frequency separation unit 1016 is quantized from the memory 1014 in accordance with the timing when the image signal captured with the shutter 1003 opened during actual imaging is output from the AZD conversion unit 1005.
- the dark current noise component is read and separated into a low frequency component and a high frequency component.
- the low frequency / high frequency separation unit 1016 separates the quantized high-frequency noise component Q of a certain pixel X into a low frequency component L and a high frequency component H according to the following equation.
- Q to Q are the quantum of each of the 15 pixels adjacent to the left side of pixel X x + 1 x + 15
- the function min2 is the function that outputs the second smallest value among the values listed in parentheses.
- the low-frequency / high-frequency separation method using Equations (5) and (6) is the dark current noise component of 16 pixels in total, 15 pixels that are adjacent to the left side of the pixel to be separated.
- the second lowest value is the low frequency component, and the dark current noise component of the pixel to be separated is subtracted from the low frequency component.
- the reason why the low frequency component is not set to the minimum value among the dark current noise components of 16 pixels is to prevent the low level signal output by the defective pixel from being the low frequency component.
- the first inverse quantization unit 1018 is a high-frequency component separated by the low-frequency / high-frequency separation unit 1016
- H is dequantized, for example, according to the following equation.
- IH H X (Vn-THn) ⁇ ⁇ ⁇ (7)
- IH is the high-frequency component of the dark current noise that has been dequantized, and this IH is the multiplier 1
- IL x is a low-frequency component of inversely quantized dark current noise.
- the second inverse quantization unit 10 20 sends the value obtained by adding the value DA of the dark current noise component at the lower limit A input from the noise distribution analysis unit 1010 to the IL to the addition unit 1028.
- the high-frequency extraction unit 1022 outputs an image output from a light shielding region force provided at the end of the image sensor 1004 with respect to an image signal captured with the shutter 1003 closed and the shutter 1003 opened.
- Signal force A high frequency component is extracted, an average value of the high frequency component is obtained, and output to the multiplication coefficient setting unit 1024.
- the high-frequency component is calculated using the equations (5) and (6) as in the low-frequency 'high-frequency separation unit 10 16.
- Multiplication coefficient setting section 1024 obtains and stores the average value of the high-frequency component in the light-shielding area calculated by high-frequency extraction section 1022 from the dark current noise component imaged in advance with shutter 1003 closed. . Further, the high frequency extraction unit 1022 obtains the average value of the high frequency components of the light shielding region in the image signal captured with the shutter 1003 opened. Of the image signals captured in this state, the image signal output from the pixel in the light shielding area is a dark current noise component in the pixel at the time when the image is captured.
- the ratio of the dark current noise component stored in advance to the average value of the high frequency component of the light shielding region is obtained, and this ratio is multiplied by the high frequency component of the dark current noise that is inversely quantized as a multiplication coefficient. Therefore, it is possible to predict a high-frequency component of dark current noise included in the captured image signal.
- the multiplication unit 1026 multiplies the high-frequency component of the dark current noise component expanded by the first inverse quantization unit by the multiplication coefficient output from the multiplication coefficient setting unit 1024, so that it is included in the image signal.
- the high frequency component of the dark current noise component is predicted.
- the adding unit 1028 generates a dark current noise component by adding the value output from the second inverse quantization unit to the predicted high frequency component. Then, the dark current noise is removed from the image signal by subtracting the dark current noise component generated by the adder 1028 from the image signal captured in the actual imaging state by the subtractor 1030.
- the image sensor 1004 captures an image with the shirt 1003 closed. At this time, the image sensor 100 Since no light is incident on 4, the image signal output from the image sensor 1004 is a negative current noise component (A).
- the dark current noise component (A) is converted into a digital signal by the AZD conversion unit 1005 and then input to the noise distribution analysis unit 1010.
- the noise distribution analysis unit 1010 obtains the magnitude distribution from the dark current noise components of all the pixels, and calculates the quantization threshold TH n and the representative value Vn based on this distribution.
- the dark current noise component (B) is imaged by the image sensor 1004 with the shutter 1003 closed again.
- This dark current noise component (B) is converted into a digital signal by the AZD conversion unit 1005 and then input to the quantization unit 1012.
- the quantization unit 1012 quantizes the dark current noise component (B) based on the threshold value Vn obtained by the noise distribution analysis unit 1010.
- the memory 1014 stores the quantized dark current noise component (B).
- the dark current noise component (B) imaged at this time is also input to the high-frequency extraction unit 1022.
- the high frequency extraction unit 1022 extracts the dark current noise component (B) force high frequency component included in the light shielding region of the image sensor 1004, and obtains an average value thereof. This average value is sent to the multiplication coefficient setting unit 1024 and stored therein.
- the operation up to here is the operation up to the stage before actually imaging the subject.
- the operation up to this point can be performed at every imaging, or can be performed when the digital camera is turned on.
- an imaging detection unit (not shown) may be provided, and when the imaging is detected, the dark current noise removing device 1001 may be instructed to perform the operation up to this stage. Since the dark current noise is easily affected by the temperature environment around the image sensor, the value of the dark current noise component can be obtained with the highest accuracy when stored immediately before imaging.
- a power-on detection unit (not shown) may be provided, and when power-on is detected, the dark current noise removing device 1001 may be instructed to perform the operation up to this stage.
- the operation up to the stage before actually capturing an image of the subject may be performed at the time of manufacturing the digital camera.
- a timer (not shown) may be provided in the digital camera so that the dark current noise removing apparatus 1001 is instructed to perform the operation up to this stage at regular intervals.
- an operation when actually imaging a subject will be described.
- an image (C) is captured by the image sensor 1004 with the shutter 1003 opened.
- the image signal (C) output from the image sensor 1004 is converted into a digital signal by the AZD conversion unit 1005 and then sent to the subtraction unit 1030 and also to the high frequency extraction unit 1022.
- the high frequency extraction unit 1022 extracts high frequency components from the image signal (C) included in the light shielding region of the image sensor 1004, obtains an average value thereof, and sends this to the multiplication coefficient setting unit 1024.
- the multiplication coefficient setting unit 1024 stores the average value of the high-frequency component of the image signal (C) contained in the light-shielding region transmitted from the high-frequency extraction unit and the high-frequency component of the dark current noise component (B) stored. The ratio with the average value is calculated and sent to the multiplication unit 1026 as a multiplication coefficient.
- the low-frequency / high-frequency separation unit 1016 performs quantization of the dark current noise component (B) stored in the memory 1014 in accordance with the timing at which the image signal (C) is output from the AZD conversion unit 1005. ) And is separated into a low frequency component (BL) and a high frequency component (BH).
- the separated high-frequency component (BH) is dequantized by the first dequantization unit 1018, and then multiplied by the multiplication coefficient obtained by the multiplication coefficient setting unit 1024 in the multiplication unit 1026, thereby generating an image.
- the high frequency component of the dark current noise component included in the signal is predicted and sent to the adder 1028.
- the separated low frequency component (BL) is dequantized by the second dequantization unit 1020 and then sent to the addition unit 1028.
- the adding unit 1028 adds dark current noise (BD) by adding the low frequency component (BL) of the dark current noise that has been dequantized and the high frequency component (BH) of the predicted dark current noise. Restore.
- the restored dark current noise component (BD) is input to the subtractor 1030, and the dark current noise component (BD) restored from the image signal (C) by the subtractor 1030 is subtracted. Thereby, the dark current noise component (BD) can be removed from the image signal (C). Then, the image signal (CD) from which the dark current noise component (BD) is removed is recorded on the recording medium 1006.
- the force shown in the example of the digital camera is not limited thereto, and the dark current noise removing device according to the embodiment of the present invention is not limited to that provided with an imaging element. Can be provided.
- the high frequency extraction unit 1022 has shown an example in which the image signal force output from the light shielding region is extracted from the high frequency components of a plurality of pixels, and the average value of the high frequency components is obtained. However, it is not limited to this, but the median or root mean square of the high frequency components of multiple pixels may be obtained.
- the high-frequency extraction unit 1022 obtains the average value of the high-frequency components of a plurality of pixels from which the dark current noise component and the image signal force are also extracted, and the multiplication coefficient setting unit 1024 calculates these values.
- the power of calculating the multiplication factor by calculating the ratio of the average value The ratio of each ratio is obtained for each pixel with respect to the dark current noise component extracted by the high frequency extraction unit 1022 and the high frequency component of the image signal.
- the average value, median value, or mean square value may be used as the multiplication coefficient.
- FIG. 10 is a diagram showing a configuration of a digital camera 2100 provided with a noise removal device 2001 according to a preferred embodiment 5 of the present invention.
- This configuration can be realized in hardware by any computer's CPU, memory, or other LSI, and in software it can be realized by a program with an encoding function loaded in the memory. Describes functional blocks realized through collaboration. Therefore, those skilled in the art will understand that these functional blocks can be realized in various forms by hardware only, software only, or a combination thereof.
- the digital camera 2100 includes the power of the noise removal device 2001, an image sensor 2002, an analog front end (AFE) 2003, an image compression device 2004, and a recording medium 2005.
- the digital camera 2100 converts the light incident on the image sensor 2002 into an electric signal, and the AFE 2003 takes out the image signal power of the image sensor 2002 and amplifies it, and then converts it into a digital signal.
- the image signal extracted by AFE2003 contains various noise components, and these noise components are removed by a noise removal device 2001 described later. Then, the image signal from which noise has been removed by the image compression apparatus 2004 is compressed and recorded on the recording medium 2005.
- the noise removal apparatus 2001 includes a fixed pattern noise removal unit 2010, a smear noise removal unit 2011, a fixed defect noise removal unit 2012, and a random noise removal unit 2013.
- FIG. 11 is a diagram showing the configuration of the fixed pattern noise removing unit 2010.
- the fixed pattern noise removal unit 2010 includes a memory 2020 and a subtraction unit 2021.
- the memory 2020 stores in advance an image signal captured by the imaging element 2002 in a state where no light is incident. Alternatively, the image signal may be compressed and stored in the memory 2020, or information associated with the image signal may be stored. An image signal picked up with no light incident corresponds to fixed pattern noise caused by dark current.
- the image is input to the fixed pattern noise removing unit 2010 via the image signal force AFE2003 picked up by the image pickup device 2002 in an actual image pickup state, that is, in a state where light is incident, the image is matched with this timing.
- the fixed pattern noise of each pixel constituting the signal is read from the memory 2020 and sent to the subtraction unit 2021. What is stored in memory 2020 If the compressed image signal or the information associated with the image signal is read out from the memory 2020, the image signal compressed for each pixel or the information associated with the image signal is read out in accordance with the above timing. It is also possible to restore the fixed pattern noise to the subtraction unit 2021 after restoring the fixed pattern noise from the signal or information. Then, the subtraction unit 2021 removes the fixed pattern noise by subtracting the fixed pattern noise from the image signal picked up with light incident thereon.
- This fixed pattern noise removing unit 2010 may take a configuration such as the dark current noise removing devices 1, 7, 8, 1001 described in the first to fourth embodiments.
- the fixed pattern noise referred to in this embodiment corresponds to the dark current noise in the first to fourth embodiments.
- FIG. 12 is a diagram showing a configuration of the smear noise removing unit 2011.
- the smear noise removal unit 201 1 includes a subtraction unit 2030, an addition unit 2031, a line memory 2032, a coefficient generation unit 2033, and a multiplication unit 2034.
- the subtracting unit 2030 also subtracts the first smear noise component Dl (n) from the image signal Sl (n) force that is continuously input in units of one row, and outputs it as an image signal S 2 (n) that does not include the smear noise component.
- the adder 2031 adds the image signal S2 (n) output from the subtractor 2030 and the cumulative addition data T (n) read from the line memory 2032 and supplies the added data to the line memory 2032.
- the line memory 2032 is reset every time the input of the image signal SI ( ⁇ ) for one screen is completed, and stores the addition data input from the addition unit 2031 for each line. As a result, the image signal S2 (n) of one screen is accumulated in each column in the adder 2031, and the accumulated data T (n) is stored in the line memory 2032.
- the coefficient generator 2033 In response to the exposure data L (m) indicating the exposure state of the image sensor 2002, the coefficient generator 2033 generates a coefficient k corresponding to the exposure period for each light receiving bit of the image sensor 2002, and a multiplier 2034 To supply.
- the multiplication unit 2034 multiplies the cumulative addition data T (n) read from the line memory 2032 by the coefficient k to generate a smear noise component Dl (n).
- the information charge accumulated in each light receiving bit of the image pickup device 2002 is transferred to the light receiving bit each time the information charge stored in the light receiving bit is transferred in the vertical direction. Amounts are accumulated sequentially. Then, since the smear noise component mixed in the process of transferring information charges is represented by the cumulative addition value, the smear noise can be removed by subtracting this value from the image signal SI (n).
- the fixed defect noise removing unit 2012 determines whether or not each pixel is a fixed defect, and if it is determined as a fixed defect, calculates a peripheral pixel force interpolation value and replaces it with that value.
- FIG. 13 is a diagram showing a configuration of the fixed defect noise removing unit 2012.
- the fixed defect noise removal unit 2012 includes a line memory 2040, a fixed defect determination unit 2041, an interpolation value calculation unit 2042, and an interpolation value replacement unit 2043.
- the line memory 2040 stores image signals that are continuously input in units of one row for each row.
- the line memory 2040 can store image signals for seven lines. When an image signal for 7 lines is stored and a new image signal is input, the image signal of the row stored at the oldest time is overwritten with the newly input image signal. From this, the contents of the line memory 2040 are updated.
- the fixed defect determination unit 2041 determines whether or not there is a fixed defect for each pixel. For example, when determining whether or not the pixel F7 shown in FIG. 14 is a fixed defect, the line memory 2 040 also reads the image signals of the peripheral pixels D5, D7, D9, F5, F9, H5, H7, and H9. Then, the maximum and minimum values of the image signals of these peripheral pixels are obtained and compared with the size of the image signal of pixel F7. As a result, when the magnitude power of the image signal of the pixel F7 is abnormally larger than the maximum value of the image signals of the peripheral pixels, or abnormally smaller than the minimum value, it is determined that the pixel F7 is a fixed defect. In other cases, it is determined that there is no fixed defect. The fixed defect determination unit 2041 performs this determination for all pixels constituting the image.
- the interpolation value calculation unit 2042 calculates an interpolation value from the values of the surrounding pixels for the pixel whose fixed defect determination unit 2041 determines whether or not it is a fixed defect.
- the interpolation value replacement unit 2043 replaces the image signal of the pixel with the interpolation value calculated by the interpolation value calculation unit 2042 and outputs it to the outside.
- the fixed defect noise is removed by outputting the image signal of the pixel as it is.
- the random noise removal unit 2013 is substantially the same as the configuration of the fixed defect noise removal unit 2012 shown in FIG. 13, but when determining the presence or absence of random noise, the maximum value of the image signal of surrounding pixels is set as the upper threshold. When the minimum value of the image signal of the surrounding pixels is set as the lower threshold, and the magnitude of the image signal of the pixel to be judged is out of the range force indicated by the upper threshold and the lower threshold It is determined that there is random noise. If it is determined that there is random noise, it is replaced with the interpolated value calculated for the pixel value of the peripheral pixel. If it is determined that there is no random noise, the image signal of that pixel is output as it is.
- the digital camera 2100 captures an image with the image sensor 2002 in a state where light is not incident.
- the image captured at this time represents a fixed pattern noise component caused by dark current.
- This image is sent to the fixed pattern noise removal unit 201 of the noise removal apparatus 2001 via the AFE 2003 and stored.
- the operation so far is the operation up to the stage before the subject is actually imaged.
- the operation up to this point can be performed at every imaging, or can be performed when the digital camera is turned on.
- an imaging detection unit (not shown) may be provided, and when the imaging is detected, the fixed pattern noise removal unit 2010 may be instructed to perform the operation up to this stage. Since the dark current noise is easily affected by the temperature environment around the image sensor, the dark current noise component value can be obtained with the highest accuracy when stored immediately before imaging.
- a power-on detection unit (not shown) may be provided, and when power-on is detected, the fixed pattern noise removal unit 2010 may be instructed to perform the operation up to this stage.
- the operation up to the stage before actually capturing an image of the subject may be performed at the time of manufacturing the digital camera.
- a timer (not shown) may be provided in the digital camera, and the fixed pattern noise removal unit 2010 may be instructed to perform the operation up to this stage at regular intervals.
- the digital camera 2100 picks up an image with the image pickup element 2002 in a state where light is incident, and the image signal is sent to the noise removing device 2001 via the AFE 2003.
- the noise removing apparatus 2001 first, the fixed pattern noise removing unit 2010 removes fixed pattern noise from the image signal, and then the smear noise removing unit 2011 removes smear noise.
- the fixed defect noise removing unit 2012 removes fixed defect noise, and the random noise removing unit 2013 removes random noise. This order eliminated multiple noises
- the image signal is compressed by the image compression apparatus 2004 and recorded on the recording medium 2005.
- the fifth embodiment is characterized in that a plurality of noises are removed in the order of fixed pattern noise, smear noise, fixed defect noise, and random noise by the noise removing device 2001. For each reason, each noise can be accurately removed.
- smear noise requires estimation of the amount of incident light as described above. Since this estimation is performed based on an image signal obtained by imaging, there is a lot of noise in this image signal, and the accuracy of estimation of the amount of incident light deteriorates. As a result, the calculation accuracy of smear noise to be removed deteriorates.
- it is difficult to accurately estimate smear noise if fixed pattern noise that is temperature dependent and increases in proportion to the exposure time is included in the image signal used for estimation. Therefore, it is desirable to remove the fixed pattern noise from the image signal before removing the smear noise.
- fixed defect noise is preferably performed on an image from which noise has been removed to some extent, since interpolation pixels for replacement with defective pixel determination are performed based on surrounding pixels. Also, depending on the amount of incident light, the smear noise may have reached a level that saturates the signal level, and the fixed defect noise will be buried. For this reason, it is desirable to remove smear noise before removing fixed defect noise.
- the random noise has a very small signal level compared to other noises, and has little effect on the removal of smear noise and fixed defect noise. However, there may be noticeable image roughness. Since this random noise removal uses a method of estimation interpolation from the characteristics of surrounding pixels, it is desirable to dispose other noise components and place them in the final stage of noise removal that is a pixel close to the true pixel value. As described above, random noise is very small compared to other noises, so the random noise removing unit may be omitted from the noise removing device 2001.
- the fixed pattern noise is removed first, followed by the smear noise, and then the fixed defect noise.
- Each noise can be removed with the highest accuracy.
- each noise including random noise can be removed with the highest accuracy by removing after removing other noise components.
- FIG. 15 is a diagram showing a configuration of a digital camera 2110 provided with a noise removal device 2006 according to a preferred embodiment 6 of the present invention.
- This noise removal device 2006 has a configuration in which an offset removal unit 2014 is added after the random noise removal unit 2013 to the noise removal device 2001 of FIG.
- the same components as those in the fifth embodiment are denoted by the same reference numerals and description thereof is omitted.
- the image signal output from the image sensor 2002 is not completely zero even if it is a black level image signal, and includes a certain offset component.
- the offset removal unit 2014 also removes this offset component from this image signal force.
- FIG. 16 is a diagram showing a configuration of the offset removing unit 2014.
- the offset removal unit 2014 includes an offset calculation unit 2050, a memory 2051, and a subtraction unit 2052.
- the offset calculation unit 2050 extracts image signals of some pixels belonging to a region (light-shielding region) where light provided at the end of the image sensor 2002 is not incident from the image signal output from the image sensor 2002. . Since the magnitude of the image signal output from the light shielding area cover is the magnitude of the black level in the image sensor 2002, the offset calculation unit 2050 uses the average value of the magnitude of the image signal as an offset component. The value is stored in the memory 2051. Then, the offset component is removed by subtracting the offset component stored in the memory 2051 from the image signal force by the subtracting unit 2052.
- the digital camera 2110 shown in FIG. 15 Based on this configuration, the operation of the digital camera 2110 shown in FIG. 15 will be described. First, similar to the digital camera 2100 shown in FIG. 10, the digital camera 2110 captures an image with the image sensor 2002 in a state where light is not incident, and this image is fixed to the noise removal device 2006 via the AFE2003. It is sent to the pattern noise removal unit 2010 and stored as a fixed pattern noise.
- the operation up to here is the operation up to the stage before the subject is actually imaged.
- the digital camera 2110 picks up an image with the image pickup element 2002 in a state where light is incident, and the image signal is sent to the noise removing device 2006 through the AFE 2003.
- the noise removal device 2006 the fixed pattern noise is first removed from the image signal by the fixed pattern noise removal unit 2010.
- the smear noise removal unit 2011 removes the smear noise.
- fixed defect noise removal unit 2012 removes fixed defect noise
- random noise removal unit 2013 removes random noise.
- the offset removal unit 2014 removes the offset component.
- the image signal from which a plurality of noises are removed in this order is compressed by the image compression device 2004 and recorded on the recording medium 2005.
- the sixth embodiment is characterized in that the offset component is removed after the fixed pattern noise is removed by the noise removing device 2006, and the offset component can be accurately removed for the following reason. Yes. That is, when calculating the offset component, the power to use the image signal of the pixel belonging to the light-shielding region of the imaging element 2002. This image signal includes a lot of fixed pattern noise caused by dark current. As described above, the fixed pattern noise has temperature dependence and increases in proportion to the exposure time. Therefore, if the offset component is calculated while including the fixed pattern noise, the calculation accuracy is deteriorated. Therefore, it is desirable to remove fixed pattern noise before removing offset components. As a result, the accuracy of removing the offset component is improved, and the image quality can be improved.
- the force shown in the example in which the offset removing unit 2014 is arranged after the random noise removing unit 2013 is not limited to this.
- the offset removing unit 2014 is not limited to the fixed pattern noise removing unit 2010. If it is arranged at a later stage, it is included in the category of the present invention.
- the force shown in the example of the digital camera is not limited thereto, and the noise removal device according to the embodiment of the present invention can be provided as long as the image pickup device is provided. .
- the present invention can be used in an apparatus that removes noise from an image sensor.
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Transforming Light Signals Into Electric Signals (AREA)
- Studio Devices (AREA)
Abstract
Description
Claims
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US11/577,787 US7924329B2 (en) | 2004-10-22 | 2005-10-18 | Noise eliminator |
JP2006543014A JP4601623B2 (ja) | 2004-10-22 | 2005-10-18 | ノイズ除去装置 |
Applications Claiming Priority (6)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2004308460 | 2004-10-22 | ||
JP2004308459 | 2004-10-22 | ||
JP2004-308459 | 2004-10-22 | ||
JP2004-308460 | 2004-10-22 | ||
JP2004-309107 | 2004-10-25 | ||
JP2004309107 | 2004-10-25 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2006043563A1 true WO2006043563A1 (ja) | 2006-04-27 |
Family
ID=36202978
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2005/019138 WO2006043563A1 (ja) | 2004-10-22 | 2005-10-18 | ノイズ除去装置 |
Country Status (4)
Country | Link |
---|---|
US (1) | US7924329B2 (ja) |
JP (1) | JP4601623B2 (ja) |
KR (1) | KR100835792B1 (ja) |
WO (1) | WO2006043563A1 (ja) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2009135787A (ja) * | 2007-11-30 | 2009-06-18 | Sony Corp | ノイズ検出装置、撮像装置、およびノイズ検出方法 |
US7924329B2 (en) | 2004-10-22 | 2011-04-12 | Sanyo Electric Co., Ltd. | Noise eliminator |
US20110102644A1 (en) * | 2007-12-10 | 2011-05-05 | Joshua Posamentier | Noise reduction in active pixel sensor arrays |
US8310567B2 (en) | 2009-11-03 | 2012-11-13 | Samsung Electronics Co., Ltd. | Methods of modeling an integrated noise in an image sensor and methods of reducing noise using the same |
CN107409182A (zh) * | 2015-03-17 | 2017-11-28 | 奥林巴斯株式会社 | 图像处理装置、图像处理方法和程序 |
JP2017216629A (ja) * | 2016-06-01 | 2017-12-07 | キヤノン株式会社 | 符号化装置、撮像装置、符号化方法、及びプログラム |
WO2018025323A1 (ja) * | 2016-08-02 | 2018-02-08 | オリンパス株式会社 | 画像処理装置、画像処理方法およびプログラム |
Families Citing this family (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8144997B1 (en) * | 2006-12-21 | 2012-03-27 | Marvell International Ltd. | Method for enhanced image decoding |
KR101464743B1 (ko) * | 2008-01-17 | 2014-11-25 | 삼성전자주식회사 | 카메라 모듈에서 신호 의존적인 잡음 추정 장치 및 방법 |
JP5366619B2 (ja) * | 2008-08-12 | 2013-12-11 | キヤノン株式会社 | 画像処理装置、撮像装置、画像処理方法、画像処理プログラム |
KR101022832B1 (ko) | 2009-11-16 | 2011-03-17 | 연세대학교 산학협력단 | Ccd센서의 스미어 열화 개선장치 및 방법 |
JP5162001B2 (ja) * | 2011-05-16 | 2013-03-13 | 株式会社東芝 | 撮像装置、撮像方法 |
KR101459034B1 (ko) * | 2013-07-11 | 2014-11-07 | 숭실대학교산학협력단 | 영상 처리 시스템 및 방법 |
DE102014224638A1 (de) | 2014-12-02 | 2016-06-02 | Olympus Soft Imaging Solutions Gmbh | Digitales Bilderfassungssystem und Verfahren zur Fehlerkorrektur in einem solchen System |
KR101521458B1 (ko) * | 2014-12-18 | 2015-05-22 | 주식회사 넥스파시스템 | 스미어 제거 장치 및 그 제거 방법 |
JP6320571B2 (ja) * | 2015-01-20 | 2018-05-09 | オリンパス株式会社 | 画像処理装置、画像処理方法およびプログラム |
JP6525756B2 (ja) * | 2015-06-15 | 2019-06-05 | キヤノン株式会社 | 放射線撮像装置及びその制御方法 |
US10249027B2 (en) * | 2016-09-22 | 2019-04-02 | Sony Corporation | Device and method for P-phase data compression |
US10193585B2 (en) * | 2017-02-17 | 2019-01-29 | SiTune Corporation | Tone removal for digital circuits |
WO2021061748A1 (en) * | 2019-09-23 | 2021-04-01 | Apple Inc. | Video compression with in-loop sub-image level controllable noise generation |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH06350930A (ja) * | 1993-06-11 | 1994-12-22 | Nec Corp | イメージセンサの暗時出力補正装置 |
JPH11298762A (ja) * | 1998-04-15 | 1999-10-29 | Toshiba Corp | 暗電流ノイズ除去装置 |
JP2000050165A (ja) * | 1998-07-27 | 2000-02-18 | Matsushita Electric Ind Co Ltd | 固体撮像装置 |
JP2001111899A (ja) * | 1999-08-02 | 2001-04-20 | Olympus Optical Co Ltd | 撮像装置及びその劣化画素信号補正方法 |
JP2003069901A (ja) * | 2001-08-23 | 2003-03-07 | Fuji Photo Film Co Ltd | 画像信号処理装置および欠陥画素補正方法 |
JP2004088306A (ja) * | 2002-08-26 | 2004-03-18 | Victor Co Of Japan Ltd | 固体撮像装置 |
JP2005295422A (ja) * | 2004-04-05 | 2005-10-20 | Matsushita Electric Ind Co Ltd | デジタルカメラ装置 |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6809763B1 (en) * | 1999-08-02 | 2004-10-26 | Olympus Optical Co., Ltd. | Image pickup apparatus and method of correcting deteriorated pixel signal thereof |
US7924329B2 (en) | 2004-10-22 | 2011-04-12 | Sanyo Electric Co., Ltd. | Noise eliminator |
-
2005
- 2005-10-18 US US11/577,787 patent/US7924329B2/en not_active Expired - Fee Related
- 2005-10-18 JP JP2006543014A patent/JP4601623B2/ja not_active Expired - Fee Related
- 2005-10-18 KR KR1020077011505A patent/KR100835792B1/ko not_active IP Right Cessation
- 2005-10-18 WO PCT/JP2005/019138 patent/WO2006043563A1/ja active Application Filing
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH06350930A (ja) * | 1993-06-11 | 1994-12-22 | Nec Corp | イメージセンサの暗時出力補正装置 |
JPH11298762A (ja) * | 1998-04-15 | 1999-10-29 | Toshiba Corp | 暗電流ノイズ除去装置 |
JP2000050165A (ja) * | 1998-07-27 | 2000-02-18 | Matsushita Electric Ind Co Ltd | 固体撮像装置 |
JP2001111899A (ja) * | 1999-08-02 | 2001-04-20 | Olympus Optical Co Ltd | 撮像装置及びその劣化画素信号補正方法 |
JP2003069901A (ja) * | 2001-08-23 | 2003-03-07 | Fuji Photo Film Co Ltd | 画像信号処理装置および欠陥画素補正方法 |
JP2004088306A (ja) * | 2002-08-26 | 2004-03-18 | Victor Co Of Japan Ltd | 固体撮像装置 |
JP2005295422A (ja) * | 2004-04-05 | 2005-10-20 | Matsushita Electric Ind Co Ltd | デジタルカメラ装置 |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7924329B2 (en) | 2004-10-22 | 2011-04-12 | Sanyo Electric Co., Ltd. | Noise eliminator |
JP2009135787A (ja) * | 2007-11-30 | 2009-06-18 | Sony Corp | ノイズ検出装置、撮像装置、およびノイズ検出方法 |
US20110102644A1 (en) * | 2007-12-10 | 2011-05-05 | Joshua Posamentier | Noise reduction in active pixel sensor arrays |
US8373778B2 (en) * | 2007-12-10 | 2013-02-12 | Intel Corporation | High speed noise detection and reduction in active pixel sensor arrays |
US8310567B2 (en) | 2009-11-03 | 2012-11-13 | Samsung Electronics Co., Ltd. | Methods of modeling an integrated noise in an image sensor and methods of reducing noise using the same |
CN107409182B (zh) * | 2015-03-17 | 2020-02-14 | 奥林巴斯株式会社 | 图像处理装置、图像处理方法和记录介质 |
CN107409182A (zh) * | 2015-03-17 | 2017-11-28 | 奥林巴斯株式会社 | 图像处理装置、图像处理方法和程序 |
JP2017216629A (ja) * | 2016-06-01 | 2017-12-07 | キヤノン株式会社 | 符号化装置、撮像装置、符号化方法、及びプログラム |
CN109565556A (zh) * | 2016-08-02 | 2019-04-02 | 奥林巴斯株式会社 | 图像处理装置、图像处理方法和程序 |
JPWO2018025323A1 (ja) * | 2016-08-02 | 2019-05-23 | オリンパス株式会社 | 画像処理装置、画像処理方法およびプログラム |
WO2018025323A1 (ja) * | 2016-08-02 | 2018-02-08 | オリンパス株式会社 | 画像処理装置、画像処理方法およびプログラム |
US10630924B2 (en) | 2016-08-02 | 2020-04-21 | Olympus Corporation | Image processing apparatus, image processing method, and non-transitory computer readable recording medium |
CN109565556B (zh) * | 2016-08-02 | 2021-02-09 | 奥林巴斯株式会社 | 图像处理装置、图像处理方法和存储介质 |
Also Published As
Publication number | Publication date |
---|---|
JP4601623B2 (ja) | 2010-12-22 |
US20090244331A1 (en) | 2009-10-01 |
US7924329B2 (en) | 2011-04-12 |
KR100835792B1 (ko) | 2008-06-05 |
KR20070063604A (ko) | 2007-06-19 |
JPWO2006043563A1 (ja) | 2008-05-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP4601623B2 (ja) | ノイズ除去装置 | |
KR101241662B1 (ko) | 화상 데이터 처리 장치, 화상 데이터 처리 방법 및 기록매체 | |
KR101464765B1 (ko) | 결합 이미지를 생성하는 방법 및 장치 | |
JP5529685B2 (ja) | 画像符号化方法、画像復号化方法、画像符号化装置及び画像復号化装置 | |
KR20070027605A (ko) | 화상 압축 처리 장치, 화상 압축 처리 방법 및 화상 압축처리 프로그램 | |
KR100690784B1 (ko) | 화질 평가를 위한 압축 영상 품질 검사 방법 | |
Lim | Characterization of noise in digital photographs for image processing | |
KR20100031599A (ko) | 촬상 장치, 촬상 제어 방법 및 촬상 제어 프로그램 | |
US20030214590A1 (en) | System and method for adaptively compensating for dark current in an image capture device | |
JP4561649B2 (ja) | 画像圧縮装置、画像圧縮プログラム及び画像圧縮方法、hdr画像生成装置、hdr画像生成プログラム及びhdr画像生成方法、並びに画像処理システム、画像処理プログラム及び画像処理方法 | |
JP2002218387A (ja) | 電子カメラ | |
JP2008011324A (ja) | ネットワークカメラ | |
JPS626273B2 (ja) | ||
Van Houten et al. | Using sensor noise to identify low resolution compressed videos from youtube | |
JP4364728B2 (ja) | 赤外線検出器の感度補正機能を有する赤外線撮像装置 | |
JP2007096607A (ja) | 画像処理装置 | |
JP4498253B2 (ja) | 画像処理装置 | |
US20060192865A1 (en) | Noise processing apparatus and image pickup apparatus | |
CN110519516B (zh) | 一种图像噪点的处理方法、装置及电子设备 | |
JPH11298762A (ja) | 暗電流ノイズ除去装置 | |
JP4253839B2 (ja) | 画像圧縮方法、画像圧縮プログラム、画像圧縮装置及びデジタルカメラ | |
JP2007142683A (ja) | デジタルスチルカメラ | |
KR100745245B1 (ko) | 휴대 단말기에서의 흔들림 움직임 보상 장치 및 방법 | |
KR100631096B1 (ko) | 주요기능이 호스트 소프트웨어에서 구현되는 비디오 카메라 | |
JP2002135658A (ja) | 撮像装置 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KM KP KR KZ LC LK LR LS LT LU LV LY MA MD MG MK MN MW MX MZ NA NG NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SM SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): BW GH GM KE LS MW MZ NA SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IS IT LT LU LV MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
WWE | Wipo information: entry into national phase |
Ref document number: 2006543014 Country of ref document: JP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 200580036077.3 Country of ref document: CN |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
WWE | Wipo information: entry into national phase |
Ref document number: 1020077011505 Country of ref document: KR |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 05795801 Country of ref document: EP Kind code of ref document: A1 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 11577787 Country of ref document: US |