JPS6135752B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention applies to each element of a silicon type image sensor such as a MOS or CCD in which elements are arranged in one or two dimensions, and each pixel is received by an individual element and converted into an electrical signal. The present invention relates to a device for digitally correcting non-uniformity such as sensitivity.

Conventionally, electron tube type sensors have been widely used to capture images, but in recent years, with advances in semiconductor technology, solid-state image sensors have entered the practical stage due to their advantages such as durability and integration. . However, a characteristic of these image sensors is that each pixel is composed of elements arranged one-dimensionally or two-dimensionally, and each element is not necessarily manufactured uniformly during the manufacturing process, resulting in non-uniformity in dark current and sensitivity. There is. Figure 1 shows an example of the output signal of a 35-element one-dimensional sensor. V ₀ (t) is the dark output for each element t, and V ₁ (t) is the bright output as well. This indicates that the dark current and bright current vary from element to element. FIG. 2 shows, as an example, the relationship between the amount of incident light I and the output voltage V of three elements t ₁ , t ₂ , and t ₃ . That is, in a silicon type image sensor, although the relationship between I and V is linear, the output V ₀ when I = O (no incident light)
(t) and the slope indicating sensitivity are different for each element. When converting such an output signal into, for example, an 8-bit digital signal using an A/D converter and performing signal processing, the configuration shown in Figure 3 is required to obtain as much information as possible from the signal. It is necessary to take That is, the output signal from the image sensor 2 is digitized via an amplifier 3 and an A/D converter 4, corrected by a sensor correction device 1, and used by a signal processing device 5. 6 is a drive circuit. The most common configuration of this sensor correction device is as shown in FIG. Here, the sensor is a 2048-element CCD. The sensor correction device 1 consists of a memory device 12 . The memory device 12 can be thought of as a collection of 2048 correction tables.
That is, one correction table corresponds to each sensor element. A/D converter 4 is included in the correction table.
The true level value (8 bits) for the digital signal (8 bit value) from the address bus 101 is stored, and when the digital signal is applied to the address bus 101, the true value is output.

Since each sensor element may have a different sensitivity from each other, 2048 correction tables are required, and each correction table corresponds to the sensor element that is the source of the signal currently input to the address bus 101. The selection is made by the drive circuit 6 giving an address. Since there are 2048 correction tables, the number of address bus lines 102 for selecting correction tables is 11.
Bits are required. The correction result is the common output line 10
3 and sent to the signal processing device 5.

In the case of the sensor correction device having the above configuration, each correction table has a capacity of 256 bytes corresponding to 8 bits of input, and 2048 correction tables are required, so a total memory device 12 of 512 Kbytes is required.
In this way, if the memory device 12 has correction values for each value of each element, the signal can be completely corrected. However, this requires a memory device with a large capacity of 4 megabits to correct the signals of the 2048-element image sensor, making the device extremely expensive.

An object of the present invention is to provide an inexpensive sensor correction device that reduces the capacity of this memory device while maintaining correction accuracy.

The present invention takes advantage of the characteristics of a silicon type image sensor and includes a dark signal memory that stores the dark signal level of each element in advance, and a subtraction circuit that subtracts the output of the dark signal memory from the output obtained by A/D converting the sensor output. , a sensitivity memory that stores the sensitivity of each element in advance, a plurality of correction memories that are addressed by the output of the sensitivity memory and accessed by each bit of the output of the subtraction circuit, and all the outputs of the plurality of correction memories. It is configured with an addition circuit that performs addition and outputs a correction result, and executes sensor correction.

The operation of the present invention will be explained below with reference to the embodiment shown in FIG. As an example, 2048 elements are used as described above.
Consider the case of CCD. It is assumed that this CCD has a non-uniformity in the dark output of each element of 5% or less of the band, and also has a non-uniformity in sensitivity of 5% or less. In other words, the output of A/D conversion takes values from 0 to 255, but since 5% of 255 is 12.75, the dark output V ₀ (t) of each element has 14 values from 0 to 13, for example. I only take it. At this time, even if a saturated light amount is given, in order to prevent the output value of each element including the dark output from exceeding 255, it is necessary to use at most 242 (= 255 - 13).
to the maximum value of the output proportional to the amount of light. The output value V ₂ (t) = V ₁ (t) − V ₀ (t) (1) of the proportional part to the saturated light amount of each element still has non-uniformity within 5% at most, so
It only takes 14 values from 229 to 242. The dark signal memory 13 stores V ₀ (t) for each element t, and the sensitivity memory 15 stores V ₂ '(t) for each element t given by equation (2). V ₂ '(t) takes a value from 0 to 13. This can be thought of as an identification number for the sensitivity of each element. Since all elements having the same V ₂ '(t) have the same sensitivity, there are only 14 different sensitivities among all 2048 elements.

V ₂ '(t)=V ₂ (t)-229...(2) The subtraction circuit 14 connects the output line 101 of the A/D converter
The data (4 bits) on the output line 130 of the dark signal memory 13 is subtracted from the data (8 bits), and the output is output to the signal line 140. Dark signal memory 1
Address No. 3 is scanned by the address bus 102 in accordance with the CCD scan. As a result, the non-uniformity of dark current was corrected. Sensitivity memory 1
5 is similarly accessed by the address bus 102, and the sensitivity identification number V ₂ '(t) which is previously measured and stored corresponding to the sensitivity of each element is transmitted to the signal line 15.
Output to 0. Each of the plurality of correction memories 16 stores the weight of each bit for each sensitivity,
Addressed by signal line 150. Signal line 14
When each bit of 0 is "1", the weight of the corresponding correction memory 16 is outputted to the output signal line 160, and when each bit is "0", o is outputted. The correction result is the result of adding all the signal lines 160 by the adding circuit 17 and is output to the signal line 103. Each sensitivity i
The weight W _ij given to each input bit j for (=V ₂ '(t)) is stored in the correction memory 16 as shown in equation (3). However, it is assumed that the value of the input voltage V ₂ is expressed as in equation (4).

Wij=255/229+ ^i2j ...(3) (Here, αj is the value of input bit j, i=0~
13, j=0 to 7) The correction result _V3 is as shown in equation (5).

For example, when sensitivity i = 0, the input voltage V ₂ is only 229 for the saturated light amount, but the corrected result V ₃ should be 255, so the correction formula is Equation (6)
become that way.

V ₃ =255/229V ₂ ...(6) At this time, for a pixel with sensitivity i = 0, V ₂ = 100
Assuming that is obtained, the true value is given by equation (6).
V ₃ =111 is obtained. (The numbers below the decimal point are rounded off.) In the present invention, this is determined by decomposing it into bits, and according to equation (3), the eight weights Woj are calculated as follows: W ₀₀ = 1.11, W ₀₁ = 2.23,
W ₀₂ = 4.45, W ₀₃ = 8.91, W ₀₄ = 17.82, W ₀₅ = 35.63,
W ₀₆ = 71.27, W ₀₇ = 142.53. Also, the input signal is V ₂ = 100, and when rewritten in binary, it becomes 01100100, so α ₀ = 0, α ₁ = 0, α ₂ = 1, α ₃ = 0 α ₄ = 0, α ₅ = 1 , α ₆ =1, α ₇ =0. Therefore, formula (5) is obtained from the above values, V ₃ = W ₀₂ + W ₀₅ + W ₀₆
= 4.45 + 35.63 + 71.27 = 111.35, which is rounded off to correctly obtain V ₃ = 111.

The fact that the correction result is the sum of the weights for each bit takes advantage of the linearity of the silicon type sensor. In addition, in the case of 8-bit data, each correction memory 1
In order to add 6 output values to have 8 bits of precision, each output signal line 160 requires 11 bits of precision, including at least 3 digits after the decimal point.
(In Figure 5, it is shown as a single line for simplicity.)
As a result, the output signal becomes a correction result with almost the same accuracy as in the case of the configuration shown in FIG. The capacity of the memory used at this time is 2048 × dark signal memory 13.
4 bits, same 2048 x 4 bits for sensitivity memory 15, 8 x 14 x 11 bits for correction memory 16, total approx.
18 kilobits, which is approximately 18 kilobits compared to the configuration shown in Figure 4.
The amount is reduced to 1/200, resulting in a significant reduction effect and lowering the cost of the equipment. In this case, a subtraction circuit and an addition circuit are used, but their delay time is not long compared to memory access, so there is no deterioration in time characteristics.

In the above embodiment, a dark signal memory and a subtraction circuit are included in the configuration in order to correct the non-uniformity of the dark signal for each element. There is a possibility that the non-uniformity becomes negligible, and in that case, the sensor correction device can be configured without including the dark signal memory and the subtraction circuit. As a result, in the above example, an additional 8 kilobits of memory can be reduced.

[Brief explanation of the drawing]

Figures 1 and 2 are illustrations of the amount of incident light and output signals of the image sensor to be corrected according to the present invention, Figure 3 is an illustration of the positioning in the image sensor system of the present invention, and Figure 4 is a general diagram. FIG. 5, which is an explanatory diagram of a sensor correction device, is an embodiment of the present invention. In the figure, 1 is a sensor correction device of the present invention, 2 is an image sensor, 3 is an amplifier, and 4 is an A/
D converter, 5 is a signal processing device, 6 is a drive circuit, 1
3 is a dark signal memory, 14 is a subtraction circuit, 15 is a sensitivity memory, 16 is a plurality of correction memories, and 17 is an addition circuit.

Claims (1)

[Scope of Claims] 1. A sensitivity memory that stores in advance the sensitivity of each element of an image sensor consisting of a plurality of elements, and an output that is addressed by the output of the sensitivity memory and that is an A/D converted output symbol of the image sensor. A sensor correction device that performs sensor correction and is comprised of a plurality of correction memories that are accessed in correspondence with each bit, and an addition circuit that adds all the outputs of the plurality of correction memories and outputs a correction result. 2. A dark signal memory that stores in advance the dark signal level of each element of an image sensor consisting of a plurality of elements, and a subtraction circuit that subtracts the output of the dark signal memory from the output obtained by A/D converting the output signal of the image sensor. , a sensitivity memory that stores the sensitivity of each element in advance, and a plurality of correction memories that are addressed by the output of the sensitivity memory and accessed correspondingly by each bit of the output of the subtraction circuit;
A sensor correction device that performs sensor correction and includes an addition circuit that adds all the outputs of the plurality of correction memories and outputs a correction result.