CN103795942B - Smear correction method of frame transfer CCD on basis of virtual reference lines - Google Patents

Abstract

The invention provides a smear correction method of a frame transfer CCD on the basis of virtual reference lines, and belongs to the technical field of photoelectric detection. In a technical route based on a reference line, dark reference lines are replaced by the virtual reference lines so as to solve the problem that the dark reference lines are prone to being polluted by stray light. When an image is read, an ROI reading technology is used so as to increase the number of the reference lines as much as possible to improve the signal to noise ratio of Smear signals; when the virtual reference lines are read, a multi-line in-chip combining technology is used so as to improve the signal to noise ratio for the largest possible, and meanwhile only one line of data need to be stored and processed additionally in the method. The method includes the specific steps of (1) frame transfer, wherein a first line which needs to be read out is transferred to the bottom end of a storage area, (2), register emptying and horizontal reading, (3) ROI image reading, (4) dumping of image lines and dark reference lines, wherein the image lines and dark reference lines do not need to be read, (5) register emptying and horizontal reading, (6) reading of the virtual reference lines after multi-line combining, (7) dark signal correction and (8) smear signal extraction and image correction.

Description

Smear Correction Method Based on Virtual Reference Line for Frame Shift CCD

technical field

The invention belongs to the technical field of photoelectric detection, and in particular relates to a Smear correction method of a frame transfer CCD.

Background technique

The Smear phenomenon is caused by the fact that the pixels in the photosensitive area of the frame transfer CCD (the effective photosensitive area in the imaging area, and the rest are dark reference pixels) are still photosensitive during the frame transfer process. It is a frame transfer CCD Inherent properties, especially in the case of strong light and short exposure time. Adding a mechanical shutter can avoid the smear phenomenon, but this is obviously a regression to the full frame transfer CCD; increasing the frequency of the frame transfer clock can weaken the smear phenomenon, but it cannot completely eliminate it.

There are currently three common correction algorithms: the first one is to estimate the location and size of the Smear based on the statistics of the whole image, which is only effective for certain types of images such as star maps, and it is difficult to apply to other images; the second is A full-frame solution algorithm based on the principle of Smear generation, it adds all rows together and multiplies them by the scale factor to obtain the Smear signal (CCD without electronic shutter) according to the mathematical process of Smear phenomenon, or starts from the first row according to the scale factor step by step Solve the ideal image after Smear correction (CCD has electronic shutter). The accuracy and signal-to-noise ratio of the Smear signal calculated by this algorithm are very high, but it requires to know the exposure time, the period of the frame transfer clock, and the generation efficiency of the Smear charge, and there must be no pixel saturation, and the entire frame must be read out ; The third is the correction algorithm based on the dark reference line, which directly uses the characteristic that the dark reference line is not sensitive to light during the exposure period, but feels the same Smear signal as the pixel in the light-sensitive area during the frame transfer process, simply After dark signal correction, the Smear signal can be directly solved. This algorithm is simple and practical, and does not require full-frame readout, but because the number of dark reference lines is usually very small, the noise of the solved Smear signal is very large, and if the illuminated area on the CCD image surface is very close to the dark reference line, In fact, these dark reference lines will be polluted by stray light and have a certain amount of exposure, and the obtained Smear signal will no longer be accurate, as shown in the upper curve in Figure 2.

Contents of the invention

The technical problem solved by the invention is to overcome the shortcomings of existing algorithms, and provide a Smear correction method based on virtual reference lines through the special design of driving timing.

The technical solution of the present invention to solve the problem:

The frame transfer CCD is based on the Smear correction method of the virtual reference line, which defines that the virtual reference behavior is after the dark reference line and outside the CCD device. It exists virtually during the exposure period, and it is transferred line by line into the CCD during the frame transfer process after the exposure. The reference row; Assuming that the imaging area has N _I rows of photosensitive pixels, N _D rows of dark reference pixels, and the storage area has N _S =N _I +N _R rows of pixels; the ROI area starts from the N _Y1 row of the imaging area (1 ≤N _Y1 ≤N _I ), there are at most N _R -N _D +(N _Y1 -1) virtual reference lines; the last three lines are discarded, and N _V =N _R -N _D +(N _Y1 -1) -3 dummy reference lines; the specific steps are as follows:

Step 1, control the frame transfer clock signal to quickly transfer the accumulated charge in the imaging area to the storage area, the number of cycles of the transfer clock is N _S + N _Y1 -1; at the same time, continuously transfer to the horizontal readout register during the frame transfer process The useless charge readout is reset or dumped;

Step 2, controlling the horizontal readout register to continuously read more than two lines, during which the readout reset pulse signal remains in a reset high level state, so as to completely remove useless charges that may remain in the horizontal readout register;

Step 3, control the row transfer clock, horizontal readout clock and readout reset pulse signal to convert the pixels in the ROI area to Read out line by line;

Step 4, controlling the row transfer clock, the horizontal readout clock and the charge dump gate control signal to dump all the useless charges between the ROI and the virtual reference row;

Step 5, execute step 2 to clear the horizontal readout register;

Step 6, control the line transfer clock to continuously transfer N _V virtual reference lines to the horizontal readout registers to realize multi-line on-chip merging, and then control the horizontal readout clock and readout reset pulse signal to read out the merged pixels one by one virtual reference line for

Step seven, perform dark signal correction on the read-out image and the virtual reference line;

Step 8: Divide the virtual reference line by N _V to obtain the Smear signal after denoising by the "multiple sampling-average" method Finally, it is subtracted line by line from the image to obtain the corrected optical signal image I _i,j .

Positive effect of the present invention:

1. The present invention chooses the technical route based on the reference line, which avoids the problems that the full-frame solution algorithm in the background technology requires full-frame readout and cannot handle saturated images. The advantage of small noise of the Smear signal is that the ROI readout technology is adopted in the method to increase the number of reference lines as much as possible and double the signal-to-noise ratio of the Smear signal.

2. The concept of "virtual reference lines" is proposed, which solves the problems that the correction algorithm based on dark reference lines is susceptible to stray light pollution, which causes the accuracy to drop, and the small number of reference lines leads to large Smear signal noise.

3. When reading the virtual reference line, the intra-slice merging technology is used to improve the signal-to-noise ratio to the greatest extent possible, and at the same time, the amount of data required for additional storage and processing by this method is only one line.

Description of drawings

Figure 1 is a frame transfer CCD pixel layout diagram with virtual reference lines.

Fig. 2 is a comparison result of the dark reference line, the virtual reference line, and the Smear signal obtained by the full-frame solving algorithm after dark level correction. Lines 1025-1027 are dark reference lines, and line 1028 is a virtual reference line. It can be seen that even the third dark reference line farthest from the photosensitive area has quite a lot of exposure, which cannot be used to calculate the Smear signal at all. However, the signal of the virtual reference line has a fairly good consistency with the Smear signal obtained by the full-frame solution algorithm.

detailed description

Below by embodiment the present invention will be further described.

As shown in FIG. 1, the virtual reference line is a reference line that virtually exists during exposure after the dark reference line and outside the CCD device. These virtual reference lines are transferred line by line into the CCD during the frame transfer process after the exposure, and then they are the same as the dark reference line, feeling and accumulating the same Smear signal as the pixel in the imaging area. Since they exist virtually during the exposure, or exposure time is zero, there is no problem of contamination by stray light. And as the number of available virtual reference lines increases, the signal-to-noise ratio of the extracted Smear signal can be doubled.

The technical solution of the present invention to solve the problem is as follows:

During frame transfer, transfer the first line of the ROI (RegionOfInterest, readout region of interest in the present invention) area in the imaging area to the bottom of the storage area, then read out the image in the ROI area line by line, and dump the ROI The useless lines between the virtual reference line and the virtual reference line are finally read out. At this time, only one added line is read out by combining multiple lines on-chip to improve the signal-to-noise ratio and reduce the amount of data.

As shown in Figure 1, it is assumed that the imaging area has N _I rows of photosensitive pixels, N _D rows of dark reference pixels, and the storage area has _NS = N _I + N _R rows of pixels. The ROI area starts from the NY1th row (1≤N _{Y1 ≤N I} ) of the imaging area, and there can be at most _{N R} -ND +( _NY1 -1) virtual reference lines. Considering that the last few lines close to the imaging area may be polluted by stray light during readout, the last 3 lines are discarded, and finally there are N _V = _{N R} -ND +( _NY1 -1)-3 virtual reference lines available.

The specific steps are described as follows:

Step 1, control the frame transfer clock signal to quickly transfer the accumulated charge in the imaging area to the storage area, the number of cycles of the transfer clock is N _S + N _Y1 -1; at the same time, continuously transfer to the horizontal readout register during the frame transfer process The useless charge readout is reset or dumped;

Step 2, controlling the horizontal readout register to continuously read more than 2 rows, during which the readout reset pulse signal remains in a reset high level state, so as to completely remove the useless charges that may remain in the horizontal readout register;

Step 3, control the row transfer clock, horizontal readout clock and readout reset pulse signal to convert the pixels in the ROI area to Read out line by line;

Step 4, controlling the row transfer clock, the horizontal readout clock and the charge dump gate control signal to dump all the useless charges between the ROI and the virtual reference row;

Step 5, execute step ② to clear the horizontal readout register;

Step 6, control the line transfer clock to continuously transfer N _V virtual reference lines to the horizontal readout registers to realize multi-line on-chip merging, and then control the horizontal readout clock and readout reset pulse signal to read out the merged pixels one by one virtual reference line for

Step seven, perform dark signal correction on the read-out image and the virtual reference line;

Step 8: Divide the virtual reference line by N _V to obtain the Smear signal after denoising by the "multiple sampling-average" method Finally, it is subtracted line by line from the image to obtain the corrected optical signal image I _i,j .

Example:

The frame transfer CCD47-20 of e2v company has an effective pixel of 1024×1024, that is, N _I =1024, N _D =3 dark reference lines, and the storage area has N _S =1024+9 lines, that is, N _R =9, ROI read The output area is lines 50-974, that is, N _Y1 =50, and finally there are N _V =9-3+(50-1)-3=52 virtual reference lines available.

CCD47-20 realizes the readout control sequence with virtual reference line as follows:

① control frame transfer clock signal with The charges accumulated in the exposure of the imaging area are quickly transferred to the storage area, and the number of cycles of the transfer clock is N _S + N _Y1 -1=1033+50-1=1082; at the same time, the horizontal readout clock is controlled with For the continuous read state, read the reset pulse signal Maintain the high level state of reset, use the "read-reset" frame transfer process to continuously transfer the useless charges to the horizontal readout register, or set the charge dumping gate clock DG high to dump the useless charges;

② Control level readout clock with For the continuous read state, read the reset pulse signal Keep it in the high level state of reset, and read more than 2 rows continuously to completely remove the useless charge that may remain in the horizontal readout register;

③ Control line transfer clock signal with Transfer a row of pixel charges to the horizontal readout register, and then control the horizontal readout clock and read reset pulse signal Read out the pixel charges one by one;

④Cycle execution step ③A total of 974-50+1=925 lines of image data are read out;

⑤ Control row transfer clock horizontal readout clock Read reset pulse signal And the charge dumping gate clock DG transfers the charges that do not need to be read out from the storage area to the horizontal readout register line by line from the 975th to 1027th lines, a total of 53 lines, and quickly dumps them;

⑥ Execute step ② to clear the horizontal readout register;

⑦Control line transfer clock signal with Continuously transfer the charge of 52 rows of virtual reference pixels to the horizontal readout register to complete the on-chip merging of virtual reference rows;

8. Execute step 3. to read out the merged virtual reference line;

⑨ Perform dark signal correction on the readout image and virtual reference line;

⑩ Divide the virtual reference line by 52 to obtain the Smear signal after noise reduction by the "multiple sampling-average" method, and finally subtract it from the image line by line to obtain the corrected optical signal image.

Claims (1)

1. Frame transfer CCD Smear correction method based on virtual reference line, which is characterized in that the virtual reference behavior is defined after the dark reference line, outside the CCD device, and it is only transferred line by line during the process of frame transfer after exposure and virtual existence during exposure. Enter the reference line that CCD begins to really exist; Suppose that imaging area has N _I row photosensitive picture element, N _D line dark reference picture element, storage area has N _S =N _I + _NR line picture element; ROI area is from imaging area Starting from line N _Y1 , 1≤N _Y1 ≤N _I , then there are at most N _R -N _D +(N _Y1 -1) virtual reference lines; the last three lines are discarded, and N _V =N _R -N _D +( N _Y1 -1)-3 virtual reference lines; the specific steps are as follows: Step 1, control the frame transfer clock signal to quickly transfer the accumulated charge in the imaging area to the storage area, the number of cycles of the transfer clock is N _S + N _Y1 -1; at the same time, continuously transfer to the horizontal readout register during the frame transfer process The useless charge readout is reset or dumped; Step 2, controlling the horizontal readout register to continuously read more than two lines, during which the readout reset pulse signal remains in a reset high level state, so as to completely remove useless charges that may remain in the horizontal readout register; Step 3, control the row transfer clock, horizontal readout clock and readout reset pulse signal to convert the pixels in the ROI area to Read out line by line; Step 4, controlling the row transfer clock, the horizontal readout clock and the charge dumping gate control signal to dump all the useless charges between the ROI region and the virtual reference row; Step 5, execute step 2 to clear the horizontal readout register; Step 6, control the line transfer clock to continuously transfer N _V virtual reference lines to the horizontal readout registers to realize multi-line on-chip merging, and then control the horizontal readout clock and readout reset pulse signal to read out the merged pixels one by one virtual reference line for Step seven, perform dark signal correction on the read-out image and the virtual reference line; Step 8: Divide the virtual reference line by N _V to obtain the Smear signal after denoising by the "multiple sampling-average" method Finally, it is subtracted line by line from the image to obtain the corrected optical signal image I _i,j .