JPS5972163A - Correcting system of sensitivity difference in image sensor - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION (a) Technical Field of the Invention The present invention relates to a method for correcting sensitivity differences between elements of a solid-state image sensor using semiconductors, and in particular improves a tegicle division circuit used for correcting sensitivity differences. This invention relates to a new signal processing method for obtaining high-quality video signals.

(b+ Background of the Technology) Recently, solid-state image sensors constructed by combining a caD (charge transfer device) with an optical sensor have made remarkable progress, and are being applied in many fields in tandem with digital image processing technology. By the way, such a solid-state image sensor is an arrangement in which a large number of light-receiving elements are integrated together with a COD multiplexer, and it is difficult to avoid paranki in the light-receiving sensitivity of each element.

This difference in sensitivity between elements is not too problematic when the level of the video signal originating from the object to be imaged is relatively large.
If the level difference between the imaged object and the background noise component is small, it may have a large effect on object identification and may impair accurate image reproduction.

(C1 Prior Art and Problems Therefore, the following method is generally considered as a way to correct the above-mentioned sensitivity difference between elements of an image sensor. In other words, the light-receiving surface of the image sensor to be corrected is All you have to do is irradiate uniform light onto the sensor, multiply it by the light reception output that exhibits a level difference depending on the sensitivity of each light receiving element, and use this as a sensitivity signal to divide the video signal during actual imaging.

The first appendix 1 (*+I″i) schematically shows the light receiving output P1-Pn for each of the famous elements when uniform light irradiation is performed on an image sensor having n light-receiving sweets, This received light output reflects the sensitivity (sensitivity) (rank) of each element, so it can be considered as sensitivity & 4R.

Eliminating the difference in sensitivity between elements means correcting the light receiving output of the AI to a uniform level of signal 48P1'-Pn' when sensitive to uniform light irradiation (Figure 1). To do this, it can be seen that ρ is obtained by dividing each of the received light outputs P, -P by its own level PI-Pn representing the sensitivity.

Here, when performing the division as described above).

Since the output of the image sensor is processed as a digital signal, a digital division circuit is used.The second configuration of this division circuit is to store the reciprocal value of the sensitivity signal in advance in memory. , it is convenient to use a method of multiplying this by the actual video signal. 2
However, if the sensitivity signal of each element is obtained from the light reception output during uniform light irradiation as described above and is expressed as a digital signal of 8 tips, then if we try to express the inverse value using the same 8 points, then Effective digits are limited, making it impossible to perform correction in a high-sensitivity region.

In other words, if you use a multiplier with N , 0 to 2550 levels (lom&) correspond to the address. At this time, the range of the reciprocal value to be output is from 254 to above, but if this reciprocal value is tabulated by the truncation method as 256 256 8 hit outputs, the sensitivity signal in the range of the input address level 129 to 265 There are now fewer bits to allocate to , and the same reciprocal value is given.

Therefore, with this method of determining the reciprocal value, it is not possible to correct the sensitivity difference in the high-sensitivity range within the range of twice as much.

The correction M degree becomes coarse. The theory is that if the bit length of the output, which is the inverse of the bit length of the input address, is multiplied by M, the bit length of the multiplication result will be increased to M pinto, and the correction accuracy will be improved by the increase in bit length. It turns out, but
There is a problem that subsequent signal processing becomes complicated.

(d) Purpose of Issuance qVi From the above-mentioned situation, the purpose of this Issuance qVi is to improve the correction #1 degree of the element sensitivity difference of the image sensor, and thus to improve the Vi-image quality. This is what we are trying to achieve.

(θ) Structure of the Invention Briefly stated, the present invention, when creating a reciprocal table of the sensitivity signal of each light-receiving element, rounds the data to 1 for a portion with a small effective bit length, that is, a small range of light-receiving output. It can be assumed that K is the valid bin of the input data)
This is characterized in that the range of M is narrowed, and as a result, the effective digits of the reciprocal value are apparently increased, thereby increasing the correction accuracy in the high sensitivity region.

(Embodiment of the Invention) To explain in more detail, when uniform light is irradiated onto the light-receiving surface of an image sensor, the light-receiving output P from each light-receiving element exhibits a sensitivity difference as shown in FIG. 1(A). , --Pfi is obtained, but if this is expressed as an 8-bit 256-level digital signal, the actual sensitivity will be O-25.
There is little variation over the level range of 5.

Therefore, in reality (ftl is 1 to 16, 1 to 3'L, or ] to 63 levels, it is safe to ignore the difference in sensitivity, and it is sufficient to consider a sensitivity range of 1 to 255 levels, for example). In other words, as shown in Figure 1 (in Al, the sensitivity signal of each light receiving element does not represent the light receiving output, but 8
The part represented by the lower 4 bits of the hint data is rounded off to one piece.

16"255 There is no problem even if the variation in the range of rehe # is expressed by 4 hints.

In this way, the effective digits of the sensitivity signal are compared to F! After compressing to 5 hints, calculate the inverse value, and carry up the output by the compressed number of digits to represent the original 8 points. This will expand the effective range of the correction signal and substantially improve the correction accuracy. can be improved.

The present invention corrects the sensitivity difference between elements of an image sensor based on the above idea, and FIG. 2 shows one embodiment of the correction circuit.

In Fig. 2, the analog light reception output P+trJ from the image sensor 1, A as a time series signal for each element]
) It enters the converter 2 and becomes 1, for example, an 8-bit digital video i. Then, a digital video signal VI1 obtained when the image sensor 1 is irradiated with uniform reference light
1 are sequentially stored in the first memory 3 as sensitivity signals corresponding to the light receiving collectors. A rounding circuit 4 is provided on the output side of the sensitivity value memory 3, which is used when a sensitivity value with a small significant digit indicating a level of 1 to 150 is read out of the eight focus signals of the sensitivity value memory 3. , the output is ignored or treated as level 16 all together.

Thus, the continuation<! Input address L of reciprocal value memory 5 of 2
i is limited to 4 bits.

Only reciprocal values corresponding to levels 16 to 255 are valid. Therefore, when reading this, the reciprocal data will not overflow even if circuit 6 is inserted and the significant digit is carried over by the 4 bits obtained by rounding the input address, and the luminance resolution will not be accurate. 8 hint reciprocal correction Yu Ji3K
You can get I.

After storing the reciprocal value data as the corrected 4FIM in the reciprocal value memory 5 from the sensitivity signal corresponding to the light reception output during uniform light irradiation as described above, when actually imaging the object, 1 time series digital data is stored. The video signal ■, and the reciprocal correction signal J8Ki read out from the reciprocal value memory 6 corresponding to the light receiving element.
The signals are combined in a digital multiplying circuit 7 with an 8×8 focus. 4 As a result 9 The output of the multiplication circuit 7 is the digital video signal v , r whose sensitivity has been corrected to correspond to the Vi light receiving element.
It becomes the benevolence that can be obtained. The reference numeral 7 in FIG.

The rounding control circuit is shown, and it is possible to set the level range for batch processing, and to adjust the number of digit shifts of the readout from the reciprocal value memory 5 in conjunction with this, allowing the correction gain to be set arbitrarily. can do. Further, the operation of each part is controlled in synchronization with the clock signal from the lock generation control circuit 8.

+f) Effects of the Invention As is clear from the above description, the present invention achieves division a function using the method of reciprocal value multiplication and corrects variations in sensitivity differences between elements of an image sensor. The second main point is to ignore the inputs and process them all at once, and to perform multiplication by increasing the effective digits of the inverse value correction signal by the number of digits in the ignored range. Therefore, the drawback of the system that the pN capability of the reciprocal correction signal is reduced particularly in the high sensitivity region is eliminated, and highly accurate correction can be performed, which is extremely advantageous for reproducing high quality images.

[Brief explanation of the drawing]

Figures 1 (A) and (B) are schematic diagrams showing a comparison of the light reception output with the sensitivity difference between the elements of the image sensor and the corrected signal output, and Figure 2 is a sensitivity difference correction according to the present invention. FIG. 2 is a block diagram showing the configuration of one embodiment of the circuit. In the figure, 1 is an image sensor, and 2-digit AD converter 8 is a tarokk generation control circuit.

Claims (1)

[Claims] (1) The output signal level of each light receiving element that appears in response to uniform light irradiation is stored in advance as an N-bit sensitivity signal, and the N bits corresponding to the reciprocal of the sensitivity signal are
A bit reciprocal correction signal is generated, and the element pair F3 is multiplied by the digital image number of N hints obtained for each light receiving element corresponding to the imaged object and the reciprocal correction signal 77 to calculate the sensitivity difference between the elements. When performing the correction, the low level portion of the sensitivity signal represented by the lower m focuses is rounded and processed, and the effective bits of the reciprocal correction signal are carried over by the rounded m bits to perform multiplication. The image sensor is characterized by a difference correction method.