KR20210142332A - Apparatus and Method for Compressing Lossless of Aurora Spectral Data - Google Patents

Abstract

According to a lossless compressing apparatus and method of aurora spectral data by an embodiment of the present invention, the method includes the steps of: allowing a central processing unit to equally divide an aurora spectrum image into a plurality of columns, and transmit each divided image to an encoder; allowing an encoder to calculate a prediction coefficient for each pixel of the segmented image; allowing the encoder to calculate the prediction value of each pixel based on the prediction coefficient; allowing the encoder to calculate a residual by subtracting the predicted value of each pixel from the original pixel value of each pixel of the divided image; and allowing the encoder to align the calculated residual, identify and remove a singular value, and encode a plurality of pieces of compressed information including the pixel value of the singular value.

Description

Apparatus and Method for Compressing Lossless of Aurora Spectral Data

The present invention relates to an apparatus and method for lossless compression of aurora spectral data.

The Northern Lights are considered to be the most beautiful wonders of nature and are colorful and constantly changing. When high-energy charged particles from space are transported to Earth by the solar wind, they are attracted to the Earth's poles by the geomagnetic field, causing molecules and atoms in the atmosphere to collide, eventually causing the aurora.

Auroras have very important research value, helping to study solar activity and its effects on Earth. Aurora studies can also provide a reference for studying other planets, since aurora is a common phenomenon for planets with magnetic fields in the solar system.

Aurora spectral data are specific hyperspectral data of the aurora spectrum, captured by pole-mounted spectrometers, and have irreplaceable research value in bridging the gap between solar activity and terrestrial evolution.

Since the exact time of occurrence of the aurora is not known, a spectrometer is required to continuously capture aurora spectral data at very high frequencies. As a result, since the amount of aurora spectral data is very large, there is a great difficulty in storage and transmission.

To solve this problem, compression of aurora spectral data is essential, and lossless compression is preferred over lossy compression. There are two main reasons for this. One is that aurora spectral images have a long-term preservation value because their scientific value is very high and the acquisition cost is quite expensive.

Various lossless compression algorithms have been proposed for compression techniques for aurora spectrum data developed so far, but transmission errors of large files are not resolved, and compression performance is poor, making it difficult to study aurora.

Korea Registered Patent No. 10-2048340

In order to solve the above problems, it is an object of the present invention to provide an apparatus and method for lossless compression of aurora spectrum data capable of improving lossless compression performance of the aurora spectrum data and minimizing the compression time.

A lossless compression method of aurora spectrum data according to a feature of the present invention for achieving the above object,

The central processing unit equally divides the aurora spectrum image into a plurality of columns, and transmits each of the divided images to an encoder;

calculating, by the encoder, a prediction coefficient for each pixel of the segmented image;

calculating, by the encoder, a predicted value of each pixel based on the prediction coefficient;

calculating, by the encoder, a residual by subtracting the predicted value of each pixel from the original pixel value of each pixel of the divided image; and

The encoder aligns the calculated residuals to identify and remove singular values, and coding a plurality of pieces of compressed information including pixel values of the singular values.

A lossless compression apparatus for aurora spectrum data according to a feature of the present invention,

a central processing unit equally dividing the aurora spectrum image into a plurality of columns and transmitting each of the divided images to an encoder;

A prediction coefficient for each pixel of the divided image is calculated, a prediction value of each pixel is calculated based on the prediction coefficient, and a prediction value of each pixel is subtracted from the original pixel value of each pixel of the divided image. The residual is calculated, and the encoder aligns the calculated residual to identify and remove a singular value, and an encoder for coding a plurality of compressed information including a pixel value of the singular value.

According to the above-described configuration, the present invention can minimize the time required to compress the aurora spectrum data, prevent transmission errors of large files due to excellent compression performance, and reduce the cost of calculating each file. there is

1 is a diagram showing the configuration of an apparatus for lossless compression of aurora spectrum data according to an embodiment of the present invention.
2 is a diagram showing the configuration of an SPCC encoder according to an embodiment of the present invention.
3 is a diagram illustrating a process of a decoder according to an embodiment of the present invention.
4 is a diagram for explaining a lossless compression method of aurora spectrum data according to an embodiment of the present invention.
5 is a diagram illustrating a result of comparing residual images after using Spec-SPCC and Spat-SPCC.
6 is a diagram illustrating an example of aurora spectrum data.
7 is a diagram illustrating a bit rate map of FIG. 6 (a) for a prediction order of 7 according to an embodiment of the present invention.

Throughout the specification, when a part "includes" a certain element, it means that other elements may be further included, rather than excluding other elements, unless otherwise stated.

Smoothing, Prediction, Classification and Coding (SPCC) for improving the lossless compression strategy of the present invention proposes a codec composed of an advanced predictive encoder and a progressive decoder. SPCC consists of smoothing, prediction, classification and coding.

The encoder uses singular value recognition and hybrid spatial spectral decorrelation.

The decoder progressively recovers original data from the compressed data of non-singular value residuals, singular value pixel values, prediction coefficients and head information.

When the total number of files in the randomly selected data set is 23,214, the codec shows the best compression performance and computational efficiency when compared to various lossless compression methods.

1 is a diagram showing the configuration of an apparatus for lossless compression of aurora spectrum data according to an embodiment of the present invention, FIG. 2 is a diagram showing the configuration of an SPCC encoder according to an embodiment of the present invention, and FIG. 3 is an embodiment of the present invention It is a diagram illustrating a process of a decoder according to an example, and FIG. 4 is a diagram for explaining a method for lossless compression of aurora spectrum data according to an embodiment of the present invention.

The apparatus 100 for lossless compression of aurora spectrum data according to an embodiment of the present invention includes an input unit 110 , a central processing unit 120 , an encoder 130 , and a decoder 140 .

The input unit 110 receives data of the aurora spectrum image from an external device.

The central processing unit 120 equally divides the aurora spectrum image into a plurality of columns, and transmits each divided image to the encoder 130 ( S100 ).

The encoder 130 is a Smoothing, Prediction, Classification and Coding (SPCC) encoder, and a Low Complexity Least Square Method (LS), which is the simplest prediction means, is applied.

The low-complexity least-squares method is unidirectional and decorrelates data along spatial or spectral dimensions.

The encoder 130 is divided into two types of spatial SPCC (Spat-SPCC) and spectral SPCC (Spec-SPCC) according to the prediction direction.

The spatial SPCC 130 includes a smoothing unit 131 , a prediction unit 132 , a classification unit 133 , and a coding unit 134 .

The spatial SPCC 130 is composed of a combination of mean shift smoothing, prediction for LS, classifying data to be compressed into four types by ORLS, and filtering the noise by compressing the classified data by a range coder.

The spectral SPCC 130 is a modified form of Spat-SPCC that reverses the position or order of the original data before smoothing.

Direct prediction is less accurate because the actual emission information is biased by noise.

Data from the original aurora spectral image is smoothed to attenuate noise effects.

The standard deviation of the data is usually large, and the best filter for Gaussian noise is the average filter.

The smoothing unit 131 smoothes the divided image using an average shift filter.

The prediction unit 132 calculates a prediction coefficient for each pixel of the segmented image using the low-complexity least squares method ( S110 ).

The prediction unit 132 calculates a prediction value of each pixel based on the prediction coefficient ( S120 ).

으로 표시되고 하기의 수학식 1에 의해 표현된다. 여기서, P(m,n)은 원본 픽셀값이고, a1 및 a2는 이전의 두 공간 픽셀에서 각각 예측되는 픽셀의 가중치 계수들이다.In fact, a 1×3 size template is used. The smoothed pixel value at a specific position (m, n) is

and is expressed by Equation 1 below. Here, P(m,n) is the original pixel value, and a1 and a2 are weight coefficients of pixels predicted in the previous two spatial pixels, respectively.

Equation 2 below is a basic equation of the low-complexity least squares method (LS).

는

공간 라인에서 대응하는 픽셀의 예측 계수이며, (m₁, n₁)에서 (m₂, n₂)까지의 이미지 블록에서 픽셀값들을

로 표시하면, 예측 계수들(

)은 수학식 2로 해결할 수 있다.Here, I' represents the minimum value between the prediction order I and the spatial location index n,

Is

It is the prediction coefficient of the corresponding pixel in the spatial line, and calculates the pixel values in the image block from _{(m 1} , n ₁ ) to (m ₂ , n _{2 ).}

Indicated by , the prediction coefficients (

) can be solved by Equation (2).

의 예측값은 수학식 2에서 계산된 예측 계수로 대체하여 결정된다.

The predicted value of is determined by substituting the prediction coefficient calculated in Equation (2).

으로 반올림되고,

는

의 반올림값 대신에

으로 계산된다.In order to reconstruct the original data without errors

rounded to ,

Is

instead of the rounded value of

is calculated as

On the other hand, pixel values of specific pixels are heavily biased by CIC (Clock Induced Charge) and become large, and consequently residuals are important.

The abnormally high-value residuals of all pixels are defined as singular values, which can degrade the overall compression performance of direct encoding.

The classifier 133 calculates a residual by subtracting the predicted value of each pixel from the original pixel value of each pixel of the divided image, and aligns the calculated residual to identify and remove the singular value ( S130 ).

This kind of singular value can be identified by the following algorithm (ORLS algorithm) as follows.

(1) Sort all absolute residuals of a specific spatial line in descending order.

는

으로 재정렬되며,

은

에 매핑되고,

은

에 매핑되고,

은

에 매핑된다.For example, if we select the nth space line,

Is

are rearranged as

silver

is mapped to

silver

is mapped to

silver

is mapped to

와

사이의 매핑 관계는 가역적이고 특이하다.index pair

Wow

The mapping relationship between them is reversible and unique.

의 정렬된 절대 잔차의 이전과 이후의 차이를 계산하고, 최대값

을 찾을 수 있다.(2)

Calculate the difference before and after the sorted absolute residuals of

can be found

보다 큰 픽셀은 특이값으로 표시된다.(3) preset absolute residual value

Larger pixels are marked with outliers.

Information to be compressed is divided into four types: Non Outlier Pixels, Outlier Pixels, Prediction Coefficients, and Header Description.

Except for the header description, an improved arithmetic coding, Range Coding (RC), is used in the last stage of each compression.

The coding unit 134 ranges-codes the remaining compressed information except for the header description by a range coder (S140).

With respect to the former two types, the residuals and the original pixel values are respective entity data to be entropy coded.

Since the described procedure of SPCC is related to a specific domain (eg, the spatial domain of Spat-SPCC), the correlation procedure of the two domains can be utilized. That is, the residual image is transformed into a one-dimensional continuous data stream through the second decorrelation dimension, divided into continuous groups, and input to the final coder instead of the sliced sub-stream.

More specifically, Spat-SPCC encodes 1024 spectral lines sequentially.

On the other hand, for the non-singular value pixels of the first two spatial lines, the original pixel values are encoded instead of the residuals because the LS must have a prediction order of 3 or greater.

Float Precision Prediction Coeffcients are quantized to improve overall compression efficiency.

The last 16 bits of each float are removed before performing the coding process directly. The left 16 bits, consisting of the sign bit, all exponent bits, and the first 7 decimal bits, are encoded.

를 넘지 않는 추가적인 절대 오차가 발생한다. 잔차에는 거의 변화가 없게 된다.for coefficients and residuals respectively

An additional absolute error that does not exceed . There is little change in the residuals.

The sign bits for all coefficients are combined together and encoded.

Exponential and decimal bits are coded in a similar manner and added successively to the compressed bit file.

Meanwhile, the headers of all custom files provide essential descriptions of key parameters for further data analysis, such as dimension size, sampling temperature, sampling interval, and exposure time. The header has a fixed size of 5760 bytes.

The progressive decoder 140 decodes a range of all compressed files. The float precision coefficients are reconstructed one by one by sequentially shifting 1, 8, and 7 bits from the decoded sign bit, exponent bit, and decimal bit to the first 16 bits of each coefficient float. The rest of the procedure is described in FIG. 3 .

In FIG. 3, numbers in squares indicate spectral position indices. For example, a square with 1 represents a pixel on the first spectral line.

The original aurora spectral data is always the predicted data plus the residuals.

Referring to the illustrated procedure for data reconstruction, it is essential to progressively decode in a single domain (Spat-S).

The prediction means are the same as the SPCC encoder 130 to ensure that the data needed to encode the current line is decoded in advance.

The progressive decoder 140 progressively decodes the original image along the n-th spatial line by using the prediction coefficients and the residual image.

5 is a diagram illustrating a result of comparing residual images after using Spec-SPCC and Spat-SPCC.

After analyzing the compression performance of two types of spatial SPCC (Spat-SPCC) and spectral SPCC (Spec-SPCC), the central processing unit 120 may select the SPCC encoder 130 having better compression performance.

As shown in Fig. 5, the experimental results of spatial SPCC (Spat-SPCC) and spectral SPCC (Spec-SPCC) both show effective compression performance, but spatial SPCC shows better performance than spectral SPCC. Able to know.

6(a) is an example of aurora spectral data generated by ZHS at UT 20:29:40 on October 19, 2015, and FIG. 6(b) is a probability distribution of (a).

Most of the pixels follow a Gaussian distribution, and the pixels are the background because there is no distribution corresponding to the light signal.

Weighting factors and prediction order affect compression performance. 7, it can be seen that the maps of bit rates for different prediction orders are similar.

7 is a diagram illustrating a bit rate map of FIG. 6 (a) for a prediction order of 7 according to an embodiment of the present invention.

, (2)

, (3)

, (4)

이다.As shown in Figure 7, given any particular prediction order, there are four cases of coefficients, for example (1)

, (2)

, (3)

, (4)

am.

The low bit rate is mapped to the hotspot area in FIG. 7 . Bits per pixel decreases as compression performance increases or decreases. When both the reciprocal of a1 and a2 are sufficiently large, it can be seen that the performance is remarkable.

For convenience of description, a coefficient corresponding to the minimum bit rate is used.

The optimal prediction order of 7 is determined from the trial with an I' range of approximately 4 to 10 for data from October 19, 2015.

In the above, the embodiment of the present invention is not implemented only through an apparatus and/or method, and may be implemented through a program for realizing a function corresponding to the configuration of the embodiment of the present invention, a recording medium in which the program is recorded, etc. And, such an implementation can be easily implemented by an expert in the technical field to which the present invention belongs from the description of the above-described embodiment.

Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improved forms of the present invention are also provided by those skilled in the art using the basic concept of the present invention as defined in the following claims. is within the scope of the right.

100: lossless compression device
110: input unit
120: central processing unit
130: encoder
140: decoder

Claims (9)

The central processing unit equally divides the aurora spectrum image into a plurality of columns, and transmits each of the divided images to an encoder;
calculating, by the encoder, a prediction coefficient for each pixel of the segmented image;
calculating, by the encoder, a predicted value of each pixel based on the prediction coefficient;
calculating, by the encoder, a residual by subtracting the predicted value of each pixel from the original pixel value of each pixel of the divided image; and
and the encoder aligns the calculated residuals to identify and remove singular values, and coding a plurality of compressed information including pixel values of the singular values. According to claim 1,
Calculating the prediction coefficient comprises:
The method of claim 1, wherein the encoder smoothes the segmented image using an average shift filter, and then calculating a prediction coefficient for each pixel of the segmented image. 3. The method of claim 2,
Calculating the prediction coefficient comprises:
The encoder lossless compression method of aurora spectrum data, characterized in that it further comprises the step of calculating the prediction coefficients using Equation 1 below, which is a low complexity least square method.
[Equation 1]

Here, I' represents the minimum value between the prediction order I and the spatial location index n,

Is

It is the prediction coefficient of the corresponding pixel in the spatial line, and calculates the pixel values in the image block from _{(m 1} , n ₁ ) to (m ₂ , n _{2 ).}

Indicated by , the prediction coefficients (

),

is the smoothed pixel value at a specific position (m, n). 4. The method of claim 3,
The smoothed pixel value is expressed by Equation 2 below and is determined by substituting the prediction coefficient of Equation 1 above.
[Equation 2]

Here, P(m,n) is the original pixel value, and a1 and a2 are weight coefficients of pixels predicted in the previous two spatial pixels, respectively. According to claim 1,
The coding step is
calculating a residual obtained by subtracting the predicted value of each pixel from the original pixel value of each pixel of the divided image, calculating the absolute residual as an absolute value, and arranging all absolute residuals of a specific spatial line in descending order;

Calculate the difference before and after the sorted absolute residuals of

finding; and
Preset absolute residual value

A method for lossless compression of aurora spectral data, further comprising the step of indicating larger pixels as singular values. According to claim 1,
The coding step is
The encoder divides the compressed information into four types of non-outlier pixels, outlier pixels, prediction coefficients, and header description, and divides the compressed information except for the header description into a range coder (Range Coder). ) Lossless compression method of aurora spectral data, characterized in that it further comprises the step of range coding by. a central processing unit equally dividing the aurora spectrum image into a plurality of columns and transmitting each of the divided images to an encoder;
A prediction coefficient for each pixel of the divided image is calculated, a prediction value of each pixel is calculated based on the prediction coefficient, and a prediction value of each pixel is subtracted from the original pixel value of each pixel of the divided image. aurora spectral data comprising an encoder that calculates a residual, the encoder aligns the calculated residual to identify and remove a singular value, and encodes a plurality of compressed information including a pixel value of the singular value Lossless compression device. 8. The method of claim 7,
The encoder is a lossless compression apparatus for aurora spectrum data, characterized in that the prediction coefficient is calculated by using Equation 1 below, which is a low complexity least square method.
[Equation 1]

Here, I' represents the minimum value between the prediction order I and the spatial location index n,

Is

It is the prediction coefficient of the corresponding pixel in the spatial line, and calculates the pixel values in the image block from _{(m 1} , n ₁ ) to (m ₂ , n _{2 ).}

Indicated by , the prediction coefficients (

),

is the smoothed pixel value at a specific position (m, n). 9. The method of claim 8,
The smoothed pixel value is expressed by Equation 2 below and is determined by substituting the prediction coefficient of Equation 1 above.
[Equation 2]

Here, P(m,n) is the original pixel value, and a1 and a2 are weight coefficients of pixels predicted in the previous two spatial pixels, respectively.

Images