KR20010073361A - Heterogeneous correction method and apparatus for video of flash burn equipment - Google Patents

Abstract

PURPOSE: A method and a device for correcting image heterogeneity in thermotropic phase equipment are provided to generate a clear screen even in a still image using a micro-scanning effect as well as being operated in real time with a digital method. CONSTITUTION: In a first step, micro-scanned frame image signals are regularized into an intermediate value of a maximum dynamic range of an A/D converter. In a second step, a gain factor of pixel unit is multiplied to the regularized image signals and an offset factor is added to the resultant value to output a pixel signal with heterogeneity being corrected by performing de-regularization. In a third step, the average value of vertical/horizontal 4 adjacent pixels is set as a reference value on the basis of the corrected pixel signals, and the difference between the reference value and a value of a current output pixel signal is output as an error value. In a fourth step, the gain factor and the offset factor are calculated in a direction that an error in a (n+1)th frame is decreased by referring the error value to the update of a correction factor.

Description

Image non-uniformity correction method and apparatus of thermal imaging equipment {HETEROGENEOUS CORRECTION METHOD AND APPARATUS FOR VIDEO OF FLASH BURN EQUIPMENT}

The present invention relates to a digital correction technology for adaptively correcting the image unevenness of the micro-scanning thermal imaging equipment, and in particular, a method for correcting image non-uniformity of the thermal imaging equipment capable of generating a clear screen without blurring of images even in a still image And to an apparatus.

Two-dimensional array infrared detectors, generally called third-generation thermal image detectors, have excellent thermal resolutions ranging from tens of thousands to hundreds of thousands of detection elements and an equal noise temperature difference of 20 mK. It causes noise and causes a decrease in observation performance. Therefore, thermal imaging equipment using a two-dimensional array infrared detector requires a correction circuit for output signal uniformity.

If the individual detection elements are perfectly linear and have time invariant characteristics, this problem can be solved by only one non-uniformity correction using a ROM. However, due to the non-linearity of the detection element, system instability, 1 / f noise, etc., the reference temperature source should be installed in the optical system and continuously corrected. However, this method has the disadvantage of not only complicated system but also bulky.

As a conventional technique proposed to solve these problems, there is a background-based nonuniformity correction method using the characteristics of the image signal of the background, such as a constant statistical technique and neural network technique.

However, in the conventional image non-uniformity correction technique using a constant statistical technique and neural network technique, since the motion of the background is a basic premise, a phenomenon that the screen is blurred and disappears in a still image is difficult, and it is difficult to implement in real time.

Accordingly, an object of the present invention is to provide a digital adaptive non-uniformity correction method and apparatus that generates a clear screen even in still images by using the micro-scanning effect to overcome the limitation of the background-based non-uniformity correction technology. have.

1 is an explanatory diagram illustrating a principle of an image nonuniformity correction method according to the present invention;

Figure 2 is an exemplary block diagram of an image non-uniformity correction device according to the present invention.

*** Description of the symbols for the main parts of the drawings ***

1: A / D converter 2: Calibration calculator

3: Gain / Offset RAM 4,6: Frame RAM

5: correction coefficient updating unit

The image non-uniformity correction method for thermal imaging equipment to achieve the object of the present invention, the first step of normalizing the micro-scanned frame video signal to the median value of the maximum dynamic range of the A / D converter; A second step of multiplying the normalized processed video signal by a gain coefficient in units of pixels, adding an offset coefficient to the resultant value, and performing denormalization to output a pixel signal having a non-uniformity corrected form; Based on the corrected pixel signal, an average value of four adjacent pixels, which are adjacent to each other (up, down, left, and right) is obtained and set as a reference value. The difference value between the reference value and the value of the currently output pixel signal is obtained, and the difference value is an error value. Outputting a third process; A fourth step of calculating the gain coefficient and the offset coefficient in a direction in which the error in the (n + 1) th frame is reduced by reflecting the error value in the correction coefficient update is performed.

FIG. 2 is a block diagram illustrating an embodiment of an image non-uniformity correction device for thermal imaging equipment to achieve an object of the present invention. As shown in FIG. 2, an A / D converter for converting a microscan image signal of a predetermined unit into digital data at high speed ( 1) and; A pipe for processing the A / D conversion of each non-uniform pixel values (x _ij (n)) each multiplied by addition unevenness correcting a gain factor (G _ij (n)) and the offset factor (O _ij (n)) in real-time A correction calculator 2 having a line structure; A gain / offset RAM 3 which receives correction factors G _ij (n) and (O _ij (n)) in predetermined units from a correction coefficient updating unit 5, which will be described later, for updating / storing the previous correction coefficients (3). )and; Storing the output pixel signal (y _ij (n)) corrected by the correction computing unit (2) and stored to be in, alternately an odd field and the corrected output pixel signal (y _ij (n)) of the even field in the Out and A frame RAM 4 having a double buffer structure for outputting; The corrected pixel signal is received from the frame RAM 4 and temporarily stored in the frame RAM 6, and the correction coefficients G _ij (n) and O _ij (n) are calculated based on the result. Is updated / stored in the gain / offset RAM 3, and a digital video signal of a predetermined unit processed in units of fields. A correction coefficient updating unit 5 for outputting a; The correction coefficient updating unit 5 comprises a frame RAM 6 that temporarily stores one frame of corrected image data required for updating the correction coefficient. The operation of the present invention configured as described above will be described in detail as follows.

First, the principle of the image non-uniformity correction method will be described with reference to FIG. 1.

When the finely scanned n-th frame video signal (x _ij (n)) is input to the non-uniformity correction network, it is normalized to the middle value (M) of the maximum dynamic range of the A / D converter, and then the gain coefficient (G) in units of pixels. _After multiplying _ij (n)) and adding the offset coefficient (O _ij (n)) to the result value, the non-uniformity corrected pixel signal y _ij (n) is output again through denormalization.

Subsequently, based on the pixel signal y _ij (n), an average value of four adjacent pixels in the up, down, left, and right directions is obtained, the average value is set as the reference value d _ij (n), and then the currently output pixel signal. The difference between the value of (y _ij (n)) and its reference value (d _ij (n)), that is, the error (e _ij (n)), is obtained and reflected in the correction coefficient update to determine the (n + 1) th frame. The gain coefficient G _ij (n) and the offset coefficient O _ij (n) are calculated in the direction of decreasing error.

Such a series of correction processes are repeatedly performed every frame, whereby the gain and offset coefficients (G _ij (n)) and (O _ij (n)) converge to stable values and the uniformity of the image is gradually improved. After a predetermined frame (for example, 200 to 300), a clear image having completely corrected unevenness is generated. The background-based nonuniformity correction algorithm based on this principle is expressed as an equation.

----------------------- (Equation 1)

-------------------------- (Equation 2)

-------- (Eq. 3)

---------------------- (Eq. 4)

Here, a is a learning coefficient that determines the convergence speed of the stability and correction coefficient of the network. As a result, the image non-uniformity correction method according to the present invention is characterized by the following two differences summarized by Equation (3) as compared with the conventional correction method.

① Normalize the digital data to the middle value (M) of the maximum dynamic range of the A / D converter to improve the stability and performance of the calibration network.

② Inverts input pixel data from output pixel data for easy hardware configuration and simplified memory structure.

Meanwhile, an embodiment of the image non-uniformity correction method as shown in FIG. 1 will be described in detail with reference to FIG. 2.

A / D converter (1) Fine scan video signal It converts to 12bit digital data at high speed.

The correction operator 2 multiplies each non-uniform pixel value x _ij (n) converted into digital data by the A / D converter 1 by a gain factor G _ij (n) and offsets the result value. As a calculator for correcting the nonuniformity by adding the coefficients (O _ij (n)), it is designed in a pipeline structure so that such correction operation can be processed in real time.

The gain / offset RAM 3 is a correction coefficient required by the correction calculator 2, that is, a gain coefficient G _ij (n), for example: , 16bit) and an offset coefficient (O _ij (n), for example: , 12 bits), and receives a correction coefficient at a predetermined cycle from the correction coefficient updating unit 5, which will be described later, and stores the correction coefficient stored in the previous step.

The frame RAM 4 is a RAM for storing the output pixel value y _ij (n) corrected by the correction operator 2, and is designed in a double buffer structure to simultaneously perform the storage and output functions. That is, each of the prescribed dose (e.g. And an even field RAM 4B having 12 bit pixels) and an even field RAM 4B when the output pixel value y _ij (n) is written to the odd field RAM 4A. The pixel value already stored in 4B is read and supplied to the correction coefficient updating unit 5, and on the contrary, the output pixel value y _ij (n) is written to the even field RAM 4B. At this time, the pixel value already stored in the odd field RAM 4A is read and supplied to the correction coefficient updater 5 side alternately in every field.

The correction coefficient update unit 5 is implemented as a digital processor as a core part that manages all signal processing in the system. The correction coefficient updating unit 5 receives the corrected pixel signal continuously read out from the frame RAM 4 and temporarily stores the correction pixel signal in the frame RAM 6, and calculates the correction coefficient using a software method. The result is then updated / stored in the gain / offset RAM 3, and the 12-bit digital video signal processed in units of fields. Outputs

Frame RAM 6 has a predetermined storage capacity (e.g., Pixel), and the correction coefficient calculating unit 5 temporarily stores data of one frame necessary for updating the correction coefficient.

As described in detail above, the present invention provides a predetermined size (eg, ) Oversample a certain number of times (e.g. 4 times) In the case of microscanning images By correcting the image unevenness by using the burned effect, it is possible to provide a clear image without blurring even in still images, and the residual noise after correction by adaptively correcting the nonlinear time-varying function characteristics of the two-dimensional array infrared detector. There is an effect that can be minimized to the thermal noise level of the detector.

Claims (7)

A first step of normalizing the finely scanned frame image signal using a normalization constant, multiplying a gain coefficient by pixel unit, adding an offset coefficient to the result value, and outputting a pixel signal corrected for non-uniformity through inverse normalization; A second step of obtaining an average value of adjacent pixels in four directions based on the corrected pixel signal, setting the average value as a reference value, and then calculating a difference value between the currently output pixel signal and an error value; The third process of calculating the gain coefficient and the offset coefficient in the direction in which the error in the (n + 1) th frame is reduced by reflecting the error value in the correction coefficient update is repeated every frame. Image non-uniformity correction method. The method of claim 1, wherein the normalization constant of the first step is a median value of the maximum dynamic range of the analog-to-digital converter. The image non-uniformity correction method of claim 1, wherein a bit shift processing learning coefficient is applied to normalize the micro-scanned frame video signal. 4. The learning coefficient of claim 3, wherein the coefficient of learning is 0.0625, 0.125, 0.25. Such as 0.5 Image non-uniformity correction method of the thermal imaging equipment characterized in that. Analog-to-digital conversion means for converting a microscan video signal of a predetermined unit into digital data at high speed and normalizing to a median of the maximum dynamic range; Correction calculation means for performing a non-uniformity correction function by multiplying each non-uniform pixel value converted into the digital data by a gain coefficient and adding an offset coefficient to the resultant value; Gain / offset storage means for receiving a gain coefficient and an offset coefficient of a predetermined unit from a correction coefficient updating means to be described later and storing the existing coefficient value in a form of updating the existing coefficient value; Frame storage means for simultaneously storing and outputting the output pixel data corrected by the correction calculating means; In calculating the gain coefficient and offset coefficient by receiving the corrected pixel data from the frame storage means, an average value of adjacent pixels is obtained, and an error value corresponding to a difference value between the average value and the value of the currently output pixel signal is obtained. and correction coefficient updating means for calculating a gain coefficient and an offset coefficient in a direction in which the error value thereof decreases in the (n + 1) th frame. The image non-uniformity correction apparatus of the thermal imaging equipment according to claim 5, wherein the correction calculating means is a hardware type calculator having a pipeline structure. The image non-uniformity correction of the thermal imaging apparatus according to claim 5, wherein the frame storage means has a double buffer structure of odd field RAM and even field RAM to alternately store and output corrected output pixel data. Device.