RU2543074C2 - Image forming apparatus - Google Patents

Abstract

FIELD: physics, video.

SUBSTANCE: invention relates to television engineering. The result is achieved due to that the image forming apparatus includes a photodetector 1, comprising two output devices, each connected to one of two identical video signal processing channels I and II, each comprising serial connection of a correlated double sampling unit 2, an amplifier unit 3 and an analogue-to-digital converter 4, the output of each of which is connected to one of the inputs of a black level referencing unit 5. Further, the invention comprises successive processing in a division unit 6, a histogram construction unit 7, a maximum determining unit 8, a correcting factor generating unit, the output of which is connected to the input of the amplifier unit in one of the processing channels.

EFFECT: providing a device which enables to perform accurate black level correction and amplification for different photodetector channels using only the captured image as a priori data.

5 dwg

Description

The invention relates to the field of television technology and can be used in the design of television cameras on photodetectors (FP) with several output devices that are used in high-definition television cameras.

For 1080p high definition television cameras, FPs have 1920 horizontal active pixels and 1080 vertical active pixels; at 25 frames per second, they require a data reading frequency of 74.25 MHz. At the same time, the maximum frequency of the output device of a matrix device with charge coupling cannot exceed 40 MHz. This limitation necessitates the use of a phase converter with two output devices, which makes it possible to read half of all pixels through one output and the second half through another output with a read frequency of 37.125 MHz. The difference in the conversion coefficients of the output devices leads to the fact that an artifact appears in the image in the form of a different brightness of the image relative to the central axis. Thus, it is necessary to compensate for the differences between the output devices of the FP.

A device for imaging is known (US Patent 6072603, Parks Cr., 2000), consisting of a phase converter with two output devices, the signals from which are fed to the input of one of the two video processing channels, each of which includes a double correlated sampling unit (BDKV), a unit amplifier (CU), analog-to-digital converter (ADC), block conversion table (BTP), and the outputs of the BTP of each channel are connected to the inputs of the memory unit (BP).

In the general case, the signal at the FP output has a nonlinear dependence on illumination, and due to the variation in the parameters of the video signal processing electric circuits, this dependence can be significant for different output devices.

The process of compensating for the unevenness of the conversion coefficients of the output devices of the FP consists in generating conversion tables obtained during the calibration process, which includes two stages - registration of the test image and generation of the conversion table.

A special test image is placed in the focus of the camera’s optical system, containing smooth gradients of brightness variation along the boundaries of the regions of the phase transition, and shooting is in progress. After the accumulation time, the process of parallel reading of the charge packets through two output devices. Then the analog signal of each output device is fed to the BDKV, which eliminates the noise of the potential setting of the floating diffusion region, flicker noise, low-frequency pickups and the clock pickup is weakened. After the signal is amplified in the control unit to the desired level and is subjected to analog-to-digital conversion. Next, digital video data is fed to the BTP, which at the calibration stage passes the input signal to the output without changes in the BP, where the signals from two channels are stored. Thus, one or more frames of the test image are captured.

After that, the coefficients are calculated in the conversion table, which is as follows. The first channel is selected as the reference, and the values for the second channel are calculated according to the least squares method:

where x is the signal value of the first channel, y is the signal value of the second channel, m is the number of quantization levels.

The coefficients a _j are a solution to the equation

All calculated coefficients are stored in memory, and the calibration process is considered complete. During normal operation, only one of two signals passes through the BTP.

The main advantage of this device is the accuracy of the degree of correction of the light-signal characteristics. The disadvantages include the need for an operator who will recalibrate with changing the subject to a test image when changing operating conditions, such as temperature, general level of lighting. This leads to the fact that such a device cannot work autonomously, which narrows the range of possible applications.

The closest set of essential features to the proposed one is the image forming device described in (US Patent 7236199 B2, Hori et al. 2007), consisting of an FP with two output devices, the signals from which are fed to the input of one of two video processing channels , each of which includes BDKV, BU, ADC. After the ADC, the signals from the two processing channels pass through the black level reference block (BPC), the neighboring pixel subtraction block (BSPP), the difference summation block (BSR), and the correction coefficient generation block (BFC). Moreover, the output of the BFKP is connected to the input of the control unit of one of the channels.

On the FP, which is divided into two equally sized areas, the image is projected. Pixels belonging to each of the areas are sequentially read through their own output device, the unevenness of the electrical parameters of which leads to noticeable artifacts in the image. The signal from each AF output first passes through the BKKV, in which the installation noise of the potential of the floating diffusion region, flicker noise, low-frequency pickups are eliminated, and the clock pickup is weakened. Then the signal is amplified in the control unit to the required level and undergoes analog-to-digital conversion. After that, the value of pixels that are closed from the light source is averaged in the BPCS. If it is possible to completely block the luminous flux, then it is possible to use the entire area of the AF, but if this is not possible, then use the values of "tempo" pixels at the edges of each of the areas of the FI. The total black level value is selected as the arithmetic average of the two values of each of the areas. After that, areas compensated for black level are counted. Then, gain compensation is made for one of the channels to the level corresponding to the reference channel. To do this, first, the difference in the values of neighboring pixels in one row is calculated, but belonging to different regions in the BVSP. Such an operation is performed in each line of the image, over several frames. In this case, the differences in the values of neighboring pixels are summed in the BSS. Based on the sign of the amount received, a decision is formed to increase or decrease the correction factor for one of the channels in the BFCC. Moreover, the step of changing the correction factor is fixed.

The main advantage of this device is the ability to compensate for differences in black level and gain for two channels based on information obtained directly from the image of the subject, which allows the system to work in automatic mode. A significant drawback is that the correction of the gain is carried out with a fixed step once per frame, which with a large ADC bit and high required accuracy can take up to several minutes.

The technical problem solved by the invention is to compensate for differences in black level and gain for two channels per image frame with high accuracy and simple implementation, allowing you to work with high-definition video in real time.

The problem is solved due to the fact that the proposed device, as well as the known one, contains a phase converter containing two output devices, each of which is connected to one of two identical video processing channels, each of which contains a serial connection of DKV, BU and ADC, and the output The ADC of each channel is connected to one of the inputs of the BPUCH, which has two inputs and two outputs, also including the BFKP, the output of which is connected to the input of the control unit of one of the channels. But unlike the known one, it contains a serial DB connection, the two inputs of which are connected to the outputs of the BPCH, BFG and BOM, the output of which is connected to the input of the BFKP.

The invention is illustrated by drawings, where in FIG. 1 shows a diagram of the proposed image forming apparatus, FIG. 2 is a schematic representation of the phase transition, in FIG. 3, a method for linking the black level for two video signal processing channels is illustrated; FIG. 4 shows the distribution of the correction coefficient along the interface between the AF regions; FIG. 5 shows a histogram of an array of correction factors.

The device operates as follows. The optical signal is projected onto the FP (1), divided into two areas and containing two output devices. In the present invention, a charge-coupled matrix device (MPS) with a size of Ν × 2M, where N is the number of rows, Μ is the number of columns in each region of the FI, is used as a phase transition. After the accumulation process, charge packets are read from the MPS in the direction indicated by the arrows in FIG. 2. Next, each three-level analog signal from two output devices passes through the corresponding channel for processing the video signal I and II (Fig. 1). The signal from the FP output in each channel enters the BDKV (2), which eliminates the noise of the potential setting of the floating diffusion region, flicker noise, low-frequency pickups, and the clock pickup is attenuated. Then the signal is amplified to the required level in the control unit (3) and passes the ADC (4).

A schematic representation of the light signal characteristic for a phase converter with two output devices is shown in FIG. 3: 1 - characteristic for channel I, 2 - characteristic for channel II. A channel with a lower black level is taken as a reference (in Fig. 3 this is channel I). Accordingly, the black level is referenced in the control unit (5) only in channel II by the value ΔU _cher , which characterizes the difference in black level in channels I and II. The black level value for each channel is calculated by averaging all the "tempo" pixels of the corresponding region on the FP. Typically, the AF has at least two “tempo” rows and two “dark” columns at the edges of the AF. Next, the difference in values is determined and one of the channels is linked to the reference level (curve 3 in Fig. 3).

Since real images have a sufficiently large spatial redundancy, we can assume with a high degree of probability that neighboring image pixels may have the same value. Based on this assumption, neighboring pixels located on the same line, but in different regions of the FP with coordinates i = 1 ... N, j = Μ, will most likely have a linear relationship

Y _{i, m} = k _{i, m} × x _{i, m} ,

where x is the pixel value of the second of the region, y is the pixel value of the second of the region, k is the linear correction coefficient. The coefficients are searched in the database (6) by dividing the pixels belonging to regions 1 and 2 with the coordinates iM, where i = l ... N (Fig. 2). Since the image also has an uneven distribution of brightness over the field, the correction coefficient depends on the coordinate, which is shown in FIG. 4. The resulting array of correction factors can be considered as a random variable. The mathematical expectation of this random variable can be estimated by the probability density. So, in BFG (7), a histogram is formed, which can be interpreted as the probability density of the correction coefficient. When constructing a histogram, only those values of the correction coefficient are taken into account that are within ± t of the value of the reference channel, where t is the tolerance for the unevenness of the conversion coefficients of the output devices of the FP, usually lying in the range from 2 to 20%. This allows you to ignore areas of the image containing high-frequency components, significantly worsening the degree of accuracy of determining the correction factor. The number of split baskets when building a histogram is selected based on the required accuracy of determining the correction factor

where R is the number of baskets. Moreover, the basket number r is associated with the correction factor as follows:

k _r = 1 - t + ε · r.

Then in BOM (8) a basket is selected that corresponds to the maximum of the histogram, and the number will determine the desired correction factor. In FIG. Figure 5 shows the histograms of the correction factor for dividing into 8, 16, 64 baskets at t = 20%. Then, in BFKP (9), the value of the desired correction factor is translated into a code corresponding to the gain / attenuation by the required number of dB, and applied to the control unit of the corresponding channel, which was not accepted as the reference one. After that, it is believed that the image formed by the FP with two output devices is corrected for black level and gain.

Achievable technical result - the device allows for fairly accurate correction of the black level and gain for different AF channels without the use of additional equipment, and using only the captured image as a priori data, which together with the small required computational costs can significantly expand the range of tasks in which high definition cameras.

Claims (1)

An image forming apparatus including a photodetector comprising two output devices, each of which is connected to one of two identical video signal processing channels, each of which contains a serial connection of a double correlated sampling unit, an amplifier unit and an analog-to-digital converter, and the output of an analog-to-digital converter each channel is connected to one of the inputs of the black level binding unit, which has two inputs that are connected to the inputs of the black level binding unit, comprising two outputs, also including a correction coefficient generating unit, the output of which is connected to the input of the amplifier unit of one of the channels, characterized in that an additional serial connection of the division unit is introduced, two inputs of which are connected to the outputs of the black level reference unit, a histogram unit, and a determination unit maximum, the output of which is connected to the input of the block forming the correction factor.

Images