KR970004171B1 - Real-time digital filter for noise elumination and gamma correction of infrared video system - Google Patents

Abstract

The noise eliminating and gamma compensating real time digital filter comprises: look-up table ROM(15-19) for storing an operation result of a weight(a) and gamma compensation coefficient(.BETA.) and for reading the operation result by using 8 bits of input image data(VDI) to output the read result as an output image data(VDO); a look-up table ROM select switch(SW13); odd frame memories(11,12) for temporarily storing odd field data and even field data of odd frames in the VDO; even frame memories(13,14) for temporarily storing odd filed data and even field data of even frames in the VDO; a frame/field read select switch(SW11) for selecting the odd and even frame memories in a read state; and a frame/field write select switch(SW12) for selecting the odd/even frame data in the VDO to temporarily store the selected data in the even/odd frame memories.

Description

Real-time Digital Filter for Noise Reduction and Gamma Compensation of Thermal Equipment

1 is a block diagram showing the basic structure of the filter,

2 is an explanatory diagram of pixel definitions on a field.

3 is a block diagram of a real-time digital filter for noise reduction and gamma compensation of the present invention.

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

11, 12: odd frame memory 13, 14: even frame memory

₁₅ ~ 19: Operation chart SW ₁₁ : Frame / field read select switch

SW ₁₂ : Frame / Field Write Selection Switch SW ₁₃ : Operation Graph Selection Switch

The present invention relates to a real-time digital filter of thermal imaging equipment, and more particularly, to a real-time digital filter for noise removal and gamma compensation of thermal imaging equipment that has effectively and comprehensively eliminated noise components and has a gamma compensation function.

In general, an image signal output from a detector of a thermal imager overlaps a noise characteristic unique to the detector and shot noise generated by a bias current. As the video signal is amplified through the shear amplifier and the main amplifier, the signal-to-noise ratio is deteriorated due to the noise characteristics of the amplifier. In particular, in a thermal system using a serial / parallel scanning method, the quantization noise is imposed on the analog / digital conversion for multiplexing the parallel analog signal into the serial signal, thereby deteriorating the video signal, thereby lowering the lower 3 to 4 bits of the digital signal. The degree is represented by noise. In addition, when a video signal is reproduced on a monitor, a gamma compensation circuit is required to linearize nonlinear electrical characteristics between the detector and the monitor.

The conventional method of removing noise by an analog filter is limited depending on the frequency bandwidth of a signal, and gamma compensation using a log amplifier has a problem with stability to temperature.

SUMMARY OF THE INVENTION In view of such a conventional problem, the present invention devised a real-time digital filter for noise removal and gamma compensation that can effectively and comprehensively remove a noise component and perform a gamma compensation function.

In the present invention, since the signal component and the noise component of the image are processed by the frame averuging technique using statistical characteristics having an uncorrelated relationship, the frequency characteristic of the signal is excellent, and also the look-up table ROM By using) to have a structure in which the calculation and the gamma compensation of the filter is processed in real time, the temperature characteristics are excellent, and the circuit can be easily implemented, which will be described in detail with reference to the accompanying drawings.

FIG. 1 is a block diagram showing the basic structure of a filter, and as shown therein, a gamma compensator 1 for receiving an input signal X _n and performing gamma compensation and amplifying the output signal of the gamma compensator 1 is shown. An amplifier (2), a delay (4) for receiving and delaying the output signal (Y _n ), an amplifier (5) for amplifying the output signal of the delay (4), and an output of the amplifier (2). And an adder 3 which adds a signal and an output signal of the amplifier 5 and outputs the output signal as Y _n .

In the above description, the input signal X _n and the output signal Y _n can be expressed by the following equation.

X _n = S _n + N _n , n = 1,2,... …

Y _n = aγ (X _n ) + bY _n-1 , Y ₀ = initial value

S _n : signal component of image, N _n : noise component of image

γ (X): αX ^β , α = 20.54, β = 0.4546 (gamma compensation coefficient)

In addition, the signal processing target image is an image having less dynamic change compared to a still image or a frame rate, that is, S ₁ = S ₂ =... … = S _n = S, uncorrelated images of signal components and noise components, i.e., images of E [SN] = 0, and images of average noise component of 0, that is, E [N] = There is a video of zero.

2 is an explanatory diagram of pixel definitions on a field. As shown in the drawing, the n-th field is defined as X _{n-2, i, j} , and the n-th field is defined as X _{n, i, j} .

In addition, the signal-to-noise ratio SNR _i of the input signal X _n may be defined as in the following equation.

E [S ² ] = Power component of the signal, E [N ² ] = Noise power component

In addition, the output signal (Y _n ) is a stable state where n is sufficiently large, and when N _n / S _n ≪1, the output signal (Y _n ) and the power component E [ε2] of the noise are expressed by the following equation. Can be.

Therefore, the signal-to-noise ratio SNR _o of the output signal Y _n can be expressed by the following equation.

It can be seen from the above equation that the signal-to-noise ratio SNR _o of the output signal Y _n can be improved according to the weight a and the gamma compensation coefficient β.

The effective averaging frame length L _e may be defined as the number of frames until the initial value Y ₀ of the output signal Y _n decreases to 0.1Y ₀ as shown in the following equation. .

L _e = -1 / log ₁₀ (1-a), with a One

The effective average frame length L _e provides important information for estimating blurring of the screen when determining the weight a in a moving image.

3 is a block diagram of a real-time digital filter for noise reduction and gamma compensation according to the present invention. As shown in FIG. 3, operation results according to a weight (a) and a gamma compensation coefficient (β) are stored and converted into a digital signal. Look-up Table ROMs (15 to 19) that use the 8-bit input image data (VDI) to be used as upper and lower addresses to read the calculation result stored therein and to output the output image data (VDO). And arithmetic expression selection switch SW ₁₃ for selecting the arithmetic expressions 15 to 19 with respect to the weight a, and odd field data and even field data for odd frames of the output image data VDO. Odd frame memories 11 and 12, which are stored and output to the operation lists 15 to 19, and odd field data and even field data for even frames of the output image data VDO are temporarily stored to perform the operation. Bromide (15-19 Frame / field read selection switch for selecting even frame memories 13 and 14 and the odd frame memories 11 and 12 and even frame memories 13 and 14 to be read. (SW ₁₁ ) and the even frame memory and the odd frame data among the output image data VDO are temporarily stored in the even frame memory 13, 14 and the odd frame memory 11, 12. It is composed of a frame / field write selection switch (SW ₁₂ ) for, and described in detail the operational effects of the present invention configured as described above.

When the 8-bit input video data VDI converted into a digital signal is input, the input video data VDI is used as the lower address and the upper address of the operation lists 15 to 19, and the operation table 15 to 19 is used. The result of the operation stored in the address is read directly and output to the output image data (VDO), resulting in a real-time operation. Here, the calculation table 15 to 19 stores the calculation result according to the weight a in five steps, and can be subdivided into several steps according to the system performance. In addition, in a system employing a microprocessor, a function of the RAM 15 to 19 may be implemented by using one RAM instead of the multiple operations 15 to 19.

In addition, operation expressions 15 to 19 having weights a of different stages as described above can be selected by operation expression selection switch SW ₁₃ . In other words, by selectively shorting the operation expression selection switch SW ₁₃ to the selection terminals a _{1 to a 5} thereof, the operation expressions 15 to 19 are selected accordingly, and the weights of the operation expressions 15 to 19 are selected. The calculation result according to (a) is output as output image data VDO.

This results in arithmetic expressions (15 to 19) for weights (a) so that blurring is less for images with little or no dynamic changes and high signal-to-noise ratios (SNR) can be obtained. Will be chosen.

In addition, the output image data VDO is selectively divided into odd frame data and even frame data by a frame / field write select switch SW ₁₂ , that is, the output image data VDO is converted into even frame data. At the time of output, the frame / field write selection switch SW ₁₂ is short-circuited to its selection terminal e ₂ so that the even frame data passes through the frame / field write selection switch SW ₁₂ and then the odd number of the even frame data. The field data is stored in the even-frame memory 13, the even-field data is stored in the even-frame memory 14, and the frame / field write selection switch SW at the time point at which the output image data VDO is output as the odd-frame data. ₁₂₎ is then through his selection terminal (o ₂₎ and a short-circuit, so that the odd-numbered frame data is the frame / field write selection switch (SW _12), the odd frame Itaconic of odd field data is stored in the odd frame memory 11, the even field data is stored in the odd frame memory 12.

In addition, the frame / field read select switch SW ₁₁ is short-circuited to the select terminal o ₁ with the frame / field write select switch SW ₁₂ shorted to its selection terminal e ₂ , so that the odd frame memory 11 12 is selected as a read state, and accordingly, odd field data and even field data of odd frames temporarily stored in the odd frame memories 11 and 12 are outputted and inputted to the operation tables 15 to 19, and the frame / The frame / field read select switch SW ₁₁ is shorted to its selection terminal e ₁ while the field write select switch SW ₁₂ is shorted to its selection terminal o ₂ , so that the even frame memories 13 and 14 Is selected as the read state, and the odd field data and even field data of the even frames temporarily stored in the even frame memories 13 and 14 are outputted and inputted to the operation lists 15 to 19. FIG.

That is, when odd field data and even field data of even frames of the output image data VDO are temporarily stored in the even frame memories 13 and 14, odd field data of odd frames temporarily stored in the odd frame memories 11 and 12. And even-numbered field data are outputted and inputted to the operational fields 15 to 19, and the odd-numbered frame memory when the odd-field data and even-field data of odd frames of the output image data VDO are temporarily stored in the odd-frame memory 11.12. Even-field data and even-field data of even frames temporarily stored in (13, 14) are outputted and inputted to the operation lists 15 to 19.

In the present invention as described above, when the gamma compensation coefficients α and β are α = 20.54 and β = 0.4546, the improvement of the signal-to-noise ratio (SNR) according to the weight a is as shown in Table 1 below.

Table 1. Improvement of SNR According to Weight (a)]

As can be seen in Table 1 above, it can be seen that the improvement in signal-to-noise ratio (SNR) increases with decreasing weight (a), and the effective average frame length increases.

As described in detail above, the present invention can effectively and comprehensively remove noise components by performing real-time calculation, perform a gamma compensation function, and simplify the circuit configuration.

Claims (3)

The calculation result is stored according to the weight (a) and the gamma compensation coefficient (β). Using the 8-bit input image data (VDI) as an address, the operation result stored therein is read and output as the output image data (VDO). Arithmetic representations 15 to 19, arithmetic representation selection switches SW ₁₃ for selecting the computational representations 15 to 19 for the weight a, and odd numbers for odd frames of the output image data VDO. Odd-field data and even-field for odd-numbered frames 11 and 12 and the output image data VDO for temporarily storing field data and even-field data and outputting them to the operational fields 15 to 19. Even frame memories 13, 14, and the odd frame memories 11, 12, and even frame memories 13, 14, which temporarily store data and output the data to the operation lists 15-19. Frame / field read selection switch (SW ₁₁ ) for selecting the read status as Frame / field write selection for selecting even frame data and odd frame data among the output image data VDO and temporarily storing them in the even frame memories 13, 14 and odd frame memories 11, 12. Real-time digital filter for noise reduction and gamma compensation of thermal imaging equipment, characterized in that consisting of a switch (SW ₁₂ ). 2. The real-time digital filter for noise elimination and gamma compensation of thermal imaging equipment according to claim 1, wherein the operation tables 15 to 19 store operation results according to weights (a) of different stages. The real-time digital for noise canceling and gamma compensation of a thermal imager according to claim 1, wherein the frame / field read select switch (SW ₁₁ ) and the frame / field write select switch (SW ₁₂ ) perform selection operations opposite to each other. Filter.