It is applied to the automatic real-time reconstruction method of multistage RTS noise of cmos image sensor
Technical field
The present invention relates to a kind of automatic real-time reconstruction method of multistage RTS noise, more particularly to one kind is applied to cmos image
The automatic real-time reconstruction method of multistage RTS noise of sensor.
Background technology
Document " V.Goiffon, G.R.Hopkinson, P.Magnan, F.Bernard, G.Rolland, and
O.Saint-Pé,"Multilevel RTS in Proton Irradiated CMOS Image Sensors
Manufactured in a Deep Submicron Technology",IEEE Trans.Nucl.Sci.,vol.56,
No.4, pp.2132-2141, Aug.2009. " discloses a kind of calculation for automatically extracting multistage RTS noise in cmos image sensor
Method.By formula
Process is filtered to original picture signal (including RTS noise), all saltus steps in primary signal are processed
For the form of triangular pulse, the filter function can realize that filtered triangular pulse height is corresponding with primary signal
Saltus step amplitude is equal, the faint saltus step of saltus step and background Gaussian noise including RTS noise signal, and wherein L represents filtering length
Degree.The standard deviation of the noise amplitude of whole primary signal is calculated, by the standard deviation sigma of whole primary signalsig, as judging that RTS makes an uproar
The threshold value of sound, when the maximum saltus step amplitude in filtered signal is more than threshold value σsig, there is RTS noise in representing primary signal, it is no
The primary signal is represented then in there is no RTS noise.
All triangular pulses and threshold value σ after it will filtersigJudge specific RTS noise saltus step, these trianglees
Amplitude change in the corresponding primary signal of pulse represents RTS noise, and other saltus steps can be considered Gaussian noise saltus step.Often
Primary signal is divided into N by the triangular pulse of RTS saltus stepsegSection, such each section of primary signal are all made an uproar not comprising RTS
Sound saltus step, by each section of standard deviation difference σsegI () is carried out averagely, obtain the standard deviation difference meansigma methodss of whole signal Gaussian noise
σgn.By meansigma methodss M of each section of amplitudesegI () is compared, if gap is more than Gaussian noise standard deviation between the two
Meansigma methodss σgn, then it is considered as two RTS noise levels, is otherwise considered as same one-level RTS noise.Amplitude after each section is processed connects
To realize the reconstruction of RTS noise signal.
There are following shortcomings in this algorithm:
1. the algorithm need etc. all primary signals terminate could start to process, in all primary signals to be calculated such as needing
Each section of standard deviation and meansigma methodss.
2. the algorithm proposes the standard deviation sigma using whole signalsigAs the threshold value for judging that RTS signals are present.Cmos image
The temperature of increase whole device of the sensor during work with the working time can rise, and cause Gaussian noise amplitude
Change, even when in the face of extreme condition, (such as radiation environment etc.) can cause device inside defect, cause RTS noise amplitude
Significantly change, and this change can cause the change of whole gaussian signal standard deviation, using the standard deviation of whole signal as threshold
Value is obviously inaccurate.
The content of the invention
Existing automatically extract in cmos image sensor the detection of multistage RTS noise algorithm noise signal and rebuild to overcome
The deficiency of credibility difference, the present invention provide a kind of automatic real-time reconstruction side of multistage RTS noise for being applied to cmos image sensor
Method.Noise signal the triangular pulse of amplitude such as be all converted to the method first, then all of negative-going pulse is changed into
All pulses are carried out trailing edge detection by direct impulse, control the mark to original noise with the trailing edge signal for detecting
Quasi- difference is calculated, the standard deviation amplitude to calculating is sampled, then by filtered triangular pulse by gain with collect
Standard deviation is compared, and using standard deviation as the threshold value for judging RTS signals, determines that triangular pulse corresponding saltus step in part is RTS
Noise signal, whenever a RTS noise saltus step is detected, resets to mean value calculation, recalculates, by the RTS for detecting
Signal controls the meansigma methodss for calculating are sampled and kept through time delay, and the average value signal for sampling is this section of RTS and makes an uproar
The amplitude of sound saltus step.The inventive method, is adopted as the threshold value for judging RTS noise automatically according to the standard deviation of real-time Gaussian noise
Collect threshold value and rebuild RTS noise signal.First, using the collection of real-time automatic signal and calculating, improve RTS noise reconstruction
Speed and efficiency, have compared with tradition Processing Algorithm and have improve precision and efficiency of detecting;Second, using real-time gaussian signal mark
Quasi- difference is compared more accurate as threshold value with the standard deviation of whole signal as threshold value, can effectively reduce temperature, space spoke
The signal standardss that degeneration of the external environment condition to the signal noise amplitude for affecting and then causing of device working environment cause such as penetrate poor
The change of (i.e. RTS noise decision threshold), can improve whole RTS noise signal detection with the credibility rebuild.
The technical solution adopted for the present invention to solve the technical problems:It is a kind of to be applied to the multistage of cmos image sensor
The automatic real-time reconstruction method of RTS noise, is characterized in comprising the following steps:
The first step, when external noise signals arrive, through formula
Filter function, complete the transformation of noise signal, the saltus step in primary signal be all converted to into etc. the three of amplitude
Angular pulse;
In formula, L represents whole filter length, aiRepresent the rising width of triangular pulse, biRepresent under triangular pulse
Drop width.Saltus step in primary signal is converted to equidirectional by the filter function, and etc. the triangular pulse of amplitude, including background is high
Faint amplitude jump in this noise.
Second step, triangular pulse is calculated by absolute value, all of negative-going pulse is changed into direct impulse;
All pulses are carried out trailing edge detection by the 3rd step, whenever a trailing edge is detected indicate that raw noise
The once generation of saltus step in signal, even in background Gaussian noise faint saltus step;
4th step, controls the standard deviation to original noise and calculates, whenever detection with the trailing edge signal for detecting
To a trailing edge signal, standard deviation is recalculated to original noise just;
5th step, the standard deviation amplitude to calculating are sampled, because standard deviation is calculated carrying out always, lead to
Cross trailing edge and detect signal standard deviation is gathered by delays time to control and be used as RTS judgment thresholds, the standard deviation Real-time Collection
Record;
6th step, filtered triangular pulse is compared with the standard deviation for collecting by gain, using standard deviation as
Judge the threshold value of RTS signals, determine that triangular pulse corresponding saltus step in part is RTS noise signal;
7th step, whenever a RTS noise saltus step is detected, resets to mean value calculation, recalculates;
8th step, by the RTS signals for detecting through time delay, controls the meansigma methodss for calculating are sampled and kept, adopts
Sample to average value signal be the amplitude of this section of RTS noise saltus step.
The invention has the beneficial effects as follows:Noise signal the triangular pulse of amplitude such as be all converted to the method first,
All of negative-going pulse is changed into into direct impulse again, all pulses are carried out with trailing edge detection, with the trailing edge letter for detecting
Number control is calculated to the standard deviation of original noise, and the standard deviation amplitude to calculating is sampled, then by filtered three
Angle pulse is compared with the standard deviation for collecting by gain, using standard deviation as the threshold value for judging RTS signals, determines part three
The corresponding saltus step of angular pulse is RTS noise signal, whenever a RTS noise saltus step is detected, mean value calculation is reset,
Recalculate, by the RTS signals for detecting through time delay, control the meansigma methodss for calculating are sampled and kept, sample
Average value signal is the amplitude of this section of RTS noise saltus step.Standard deviation conduct of the inventive method according to real-time Gaussian noise
Judge the threshold value of RTS noise, automatic data collection threshold value and rebuild RTS noise signal.First, using real-time automatic signal collection with
Calculate, improve the speed and efficiency of RTS noise reconstruction, compared with tradition Processing Algorithm and improve precision and efficiency of detecting;
Second, using real-time gaussian signal standard deviation as threshold value, compare more smart as threshold value with the standard deviation of whole signal
Really, the signal noise that affect and then cause of the external environment conditions such as temperature, space radiation to device working environment can effectively be reduced
The change of the signal standardss poor (i.e. RTS noise decision threshold) that the degeneration of amplitude causes, improves whole RTS noise signal detection
With the credibility rebuild.
With reference to the accompanying drawings and detailed description the present invention is elaborated.
Description of the drawings
Fig. 1 is that the present invention is applied to the automatic real-time reconstruction method of multistage RTS noise of cmos image sensor in signal
Schematic diagram after RTS noise system reconstructing.
Specific embodiment
With reference to Fig. 1.The present invention is applied to the automatic real-time reconstruction method of multistage RTS noise of cmos image sensor and specifically walks
It is rapid as follows:
The first step, when external noise signals arrive, through formula
Filter function, complete the transformation of noise signal, the saltus step in primary signal be all converted to into etc. the three of amplitude
Angular pulse;
In formula, L represents whole filter length, aiRepresent the rising width of triangular pulse, biRepresent under triangular pulse
Drop width.Saltus step in primary signal is converted to equidirectional by the filter function, and etc. the triangular pulse of amplitude, including background is high
Faint amplitude jump in this noise.
Second step, triangular pulse is calculated by absolute value, all of negative-going pulse is changed into direct impulse;
All pulses are carried out trailing edge detection by the 3rd step, whenever a trailing edge is detected indicate that raw noise
The once generation of saltus step in signal, even in background Gaussian noise faint saltus step;
4th step, controls the standard deviation to original noise and calculates, whenever detection with the trailing edge signal for detecting
To a trailing edge signal, standard deviation is recalculated to original noise just;
5th step, the standard deviation amplitude to calculating are sampled, because standard deviation is calculated carrying out always, lead to
Cross trailing edge and detect signal standard deviation is gathered by delays time to control and be used as RTS judgment thresholds, the standard deviation Real-time Collection
Record;
6th step, filtered triangular pulse is compared with the standard deviation for collecting by certain gain, by standard
As the threshold value for judging RTS signals, difference determines that triangular pulse corresponding saltus step in part is RTS noise signal;
7th step, whenever a RTS noise saltus step is detected, resets to mean value calculation, recalculates;
8th step, by the RTS signals for detecting through certain time-delay, controls the meansigma methodss for calculating are sampled and protected
Hold, the average value signal for sampling is the amplitude of this section of RTS noise saltus step.
Fig. 1 is the multistage RTS noise automatic detection and reconstructing system that the inventive method is applied to cmos image sensor.Bag
Containing filtering part, threshold level produces part and sampling holding part.Wherein filtering part is comprising filter function module and definitely
Value computing module;Threshold level generating unit point includes gain module, declines detection module, and standard deviation computing module, high level are adopted
Egf block and time delay module;Sampling holding part includes comparator module, mean value calculation module, rising edge sampling module and
Time delay module.All saltus steps are filtered into the triangular pulse of the amplitude such as equidirectional by filter function module by primary signal, then
All triangular pulses are arranged as pulse upwards by absolute value block.Filtered triangular pulse signal is carried out down
Drop often detects a trailing edge and means that primary signal there occurs a saltus step along detection, and trailing edge is compared to rising edge
The saltus step of signal has been completed, and amplitude is relatively more stable and reliable, is detected trailing edge and is controlled follow-up standard deviation calculating mould
Block resets and recalculates, and the standard deviation that collection a period of time calculates in the effective module of high level judges RTS skip signals the most
Threshold value.Filtered signal and standard deviation threshold method are compared, and this saltus step are assert for RTS noise, product more than threshold value
Raw Pulse Width Control rising edge sampling module gathers corresponding meansigma methodss as the corresponding saltus step amplitude of this grade of RTS noise.Otherwise
Regard as Gaussian noise saltus step.Wherein filtering part realizes the triangular pulse that saltus step in primary signal is filtered into different amplitudes
Signal, threshold level generation part are realized and calculate Gaussian noise standard deviation conduct in primary signal in real time according to filtered signal
Threshold value, holding part of sampling compare to determine out RTS saltus steps and gather corresponding RTS according to filtered signal and threshold level
The amplitude of saltus step, completes the reconstruction of RTS signals.
The inventive method is according to the standard deviation of real-time Gaussian noise as the threshold value for judging RTS noise, automatic data collection threshold
Value and rebuild RTS noise signal.First, using the collection of real-time automatic signal and calculating, improve the speed of RTS noise reconstruction
And efficiency, compare with tradition Processing Algorithm and improve precision and efficiency of detecting;Second, using real-time gaussian signal standard deviation
As threshold value, compare more accurate as threshold value with the standard deviation of whole signal, can effectively reduce temperature, space radiation etc.
The signal standardss that degeneration of the external environment condition to the signal noise amplitude for affecting and then causing of device working environment causes are poor (i.e.
RTS noise decision threshold) change, improve whole RTS noise signal detection with rebuild credibility.