CN119006330B - A method for infrared image denoising based on SGM-delta - Google Patents

Abstract

The invention discloses an infrared image denoising method based on sgm-delta, which comprises the following steps of S1, determining a point set capable of participating in filtering in a neighborhood, S2, carrying out sgm-delta quantization coding on the point set, and S3, calculating a denoising result through a coding value. The invention carries out image denoising based on sgm-delta noise shaping principle, and as the bandwidth is widened, the noise energy is equally divided into the whole bandwidth range, so that the bottom noise is reduced, and the signal to noise ratio is improved.

Description

Sgm-delta-based infrared image denoising method

Technical Field

The invention relates to the technical field of infrared image processing, in particular to an infrared image denoising method based on sgm-delta.

Background

The infrared image is influenced by detector non-uniformity and dark current noise of a reading circuit, and the image picture often remains more noise. Common denoising methods include kernel-based airspace methods (bilateral filtering, guided filtering, nlm and the like), multi-block matching-based methods (dnr, bm3d, block svd and the like), and transform domain-based (wavelet base, frequency domain denoising and the like), wherein the kernel-based airspace methods act on the center point of a current block through kernel operation, the denoising effect can leave marks of airspace calculation (excessively strong noise points can be smoothed into 'protrusions'), the defects are overcome in the multi-block matching-based methods, more reference information is provided for denoising through increasing the number of blocks, the block superposition form acts on each point of the current block, but the algorithm calculation amount and the hardware implementation complexity are increased by the multi-block-based methods, and the computational complexity and the memory resource overhead are high in the wavelet base, the frequency domain based on transform domain algorithm.

Disclosure of Invention

The invention provides more reference information for denoising based on the advantages of a plurality of matching denoising ideas, considers the complexity reduction of an algorithm, and provides a novel infrared image denoising ideas and an image denoising method based on sgm-delta noise shaping principle.

The invention aims at adopting the following technical scheme to realize the infrared image denoising method based on sgm-delta, which comprises the following steps:

S1, determining a point set which can participate in filtering in a neighborhood;

S2, carrying out sgm-delta quantization coding on the point set;

S3, calculating a denoising result through the coding value.

Further, step S1 is specifically to judge whether another point coordinate (i ', j') in the neighborhood of the coordinate (i, j) point satisfies a preset condition, if so, the value of the point (i ', j') is classified into a point set { Y _ij };

wherein, the preset conditions are:

;

In the formula, Representing image values at coordinates (i, j) on infrared image Ysrc; Representing image values at coordinates (i ', j') on the infrared image Ysrc; Representing a threshold value, related to the noise level.

Further, step S2 includes the following sub-steps:

S21, calculating a filtering reference value Vref+ and Vref-;

S22, traversing the point set { Y _ij }, and calculating a coded value Bcode.

Further, step S21 includes the following sub-steps:

s211, calculating a mean value ymean of the point set { Y _ij }:

wherein n is the number of elements in the point set { Y _ij };

S212, calculating a weight average value Vref:

Wherein w represents a weight inversely proportional to the neighborhood variance var;

S213, calculating a filtering reference value:

;

Wherein vd represents Is a fluctuation level of (c).

Further, step S22 includes the following sub-steps:

S221, initializing variable values;

S222, calculating an output value of the comparator;

s223, calculating an output value of the integrator;

s224, judging the sign of the output value of the integrator and accumulating the code stream;

s225, updating a reference value of the comparator;

s226, returning to the step S222 until each element in the point set is traversed.

Further, the initialization variable values include one or more of a counter k zero, a negative side reference value vdac _k zero, an integrator accumulated value sum_k zero, and a stream accumulated value Bcode zero.

Further, the step S3 specifically includes calculating a denoising result from the code value Bcode:

;

Wherein yft is the denoising result.

Further, the neighborhood includes a spatial domain, or a temporal domain, or a time-spatial domain.

The invention has the beneficial effects that the image denoising is carried out based on sgm-delta noise shaping principle, and the noise energy is equally divided into the whole bandwidth range due to the widening of the bandwidth, so that the background noise is reduced, and the signal to noise ratio is improved.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to the structures shown in these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a flow chart of the present invention;

FIG. 2 is a flowchart of quantization encoding of a point set;

FIG. 3 is a flowchart for calculating a filtered reference value;

FIG. 4 is a flowchart for calculating code values;

Fig. 5 is a block diagram of a sigma-delta modulator.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.

It should be noted that like reference numerals and letters refer to like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures.

Some embodiments of the present invention are described in detail below with reference to the accompanying drawings. The following embodiments and features of the embodiments may be combined with each other without conflict.

Referring to FIG. 1, an infrared image denoising method based on sgm-delta comprises the following steps:

S1, determining a point set which can participate in filtering in a neighborhood;

S2, carrying out sgm-delta quantization coding on the point set;

S3, calculating a denoising result through the coding value.

The sgm-delta noise shaping principle is a technology for utilizing high-precision sampling to move quantization noise to a high-frequency region and filtering noise by low-pass filtering, and as the bandwidth is widened, the noise energy is equally divided into the whole bandwidth range, so that the bottom noise is reduced, and the signal to noise ratio is improved. The invention refers to a multi-block matching denoising idea, and replaces high-precision sampling in the time domain by utilizing multi-block information in the space domain, thereby realizing an infrared image denoising method based on sgm-delta.

Specifically, sgm-delta noise shaping techniques are techniques such as oversampling of noise by sigma-delta analog-to-digital converters in the vicinity of digital electronics, analog electronics, noise shaping, low pass filtering, and the like. The invention uses the principle of the sigma-delta analog-to-digital converter and applies the principle to denoising of infrared images. The invention also includes over-sampling, noise shaping, and low-pass filtering. The differences are:

In the oversampling link, the sigma-delta analog-to-digital converter realizes oversampling in a frequency domain by increasing the sampling rate, and the invention realizes 'oversampling' by using neighborhood multiple points in a airspace. First, consider the frequency domain transfer function of a conventional multi-bit ADC when the input signal is a sine wave. The input is sampled at a frequency Fs. According to the nyquist theorem, fs must be at least twice the bandwidth of the input signal. The FFT analysis result output from the digital, when the signal contains a single tone and a large amount of random noise, extends from dc to Fs/2. These noise, called quantization noise, can be considered for the result as an ADC input being a continuous signal with infinite possible states, but the digital output being a discrete function, the number of different states depending on the resolution of the converter. Therefore, the conversion from analog to digital loses some information and introduces a degree of distortion in the signal. The magnitude of the error is random, with a maximum of + -LSB. In order to improve the signal-to-noise ratio and reduce the noise floor noise, the traditional method is to increase the bit width of the ADC, thus increasing the bit width and the power consumption of the whole system, and the oversampling technology is different, so that the sampling frequency is improved (by adopting an oversampling factor k to achieve kFs), the noise floor noise can be reduced, and the bit width is only increased by 1 bit, so that the sigma-delta converter realizes a wider dynamic range with the ADC with lower resolution.

In the noise shaping link, the threshold value of the comparator of the sigma-delta analog-to-digital converter is a fixed value and is related to the signal amplitude range, and the invention is to adaptively calculate the two thresholds of the comparator through image local information. Taking a block diagram of a first order sigma-delta modulator as an example (as shown in fig. 5), it comprises a differential amplifier, an integrator and a comparator, and a feedback loop comprising a 1-bit DAC. (the DAC is a simple switch that connects the negative input of the differential amplifier to either a positive or negative reference voltage). The purpose of the feedback DAC is to maintain the average output of the integrator close to the reference level of the comparator. The density of the modulator output "1" is proportional to the input signal. The comparator generates a large number of "1" s when the input increases, and vice versa when the input decreases. By summing the error voltages, the integrator is a low pass filter for the input signal and a high pass filter for the quantization noise. Therefore, most quantization noise is shifted to higher frequencies. Oversampling not only alters the total noise power, but also its distribution.

In the low-pass filtering step, the sigma-delta analog-to-digital converter uses a decimation filter to achieve low-pass filtering, while the invention uses (but is not limited to) an averaging filter to achieve low-pass filtering. The output of the sigma-delta modulator is a 1-bit data stream and the sampling rate can reach the megahertz range. The purpose of the digital and decimated filters is to extract information from the data stream, reducing the data rate to a more useful value. In a sigma-delta ADC, a digital filter averages the 1-bit data stream, improving ADC resolution, and filtering out-of-band quantization noise.

The method comprises the steps of taking points in the neighborhood of the current point, wherein the neighborhood space is a block with a large radius (the neighborhood radius of a general denoising algorithm is 3*3 or 5*5, the radius is smaller, the neighborhood radius of the method is 16 x 16,32 x 32 and the like), and the purpose of the method is to realize oversampling by using space information, which corresponds to the step S1 of the method, the noise shaping part corresponds to the step S2 of the method, and the low-pass filtering part corresponds to the step S3 of the method.

It should be noted that, in the present invention, on the infrared image Ysrc with height H and width W, the infrared image Ysrc is 14bit original image data after non-uniform correction, a denoising process is performed on each point on the infrared image Ysrc, the image value at the coordinate (i, j) on the infrared image Ysrc is Ysrc (i, j), the neighborhood of the point is a square area of d×d, the coordinate of the top left vertex of the neighborhood is (i-r, j-r), the coordinate of the bottom right vertex of the neighborhood is (i+r, j+r), d=r+r+1, where r is the neighborhood radius, and D is preferably a neighborhood diameter, and D is preferably 65.

Further, in step S1, the coordinates of a point in the neighborhood of the point of coordinates (i, j) are set to (i ', j'), and if the following condition is satisfied, the value of the point (i ', j') is included in the point set { Y _ij }

The conditions are as follows:

;

In the formula, Representing image values at coordinates (i, j) on infrared image Ysrc; Representing image values at coordinates (i ', j') on the infrared image Ysrc; the representation threshold, which is related to the noise level, may be set by the user, preferably 30.

Referring to fig. 2, step S2 includes the sub-steps of:

S21, calculating a filtering reference value Vref+ and Vref-;

S22, traversing the point set { Y _ij }, and calculating a coded value Bcode.

Referring to fig. 3, step S21 includes the sub-steps of:

s211, calculating a mean value ymean of the point set { Y _ij }:

wherein n is the number of elements in the point set { Y _ij };

S212, calculating a weight average value Vref:

wherein w represents a weight inversely proportional to the neighborhood variance var, where the local variance is large is achieved, vref tends to the image value at the (i, j) point Where the local variance is small, vref tends to the neighborhood mean;

The relationship of weights to neighborhood variances may be, but is not limited to:

Where eps is a constant set by the user, preferably 100;

the calculation formula of the local variance is:

;

S213, calculating a filtering reference value:

;

Wherein vd represents Can be set by the user, preferably 20.

Referring to fig. 4, step S22 includes the sub-steps of:

S221, initializing variable values;

S222, calculating an output value of the comparator;

s223, calculating an output value of the integrator;

s224, judging the sign of the output value of the integrator and accumulating the code stream;

s225, updating a reference value of the comparator;

s226, returning to the step S222 until each element in the point set is traversed.

The initialization variable value comprises one or more of a counter k, a negative terminal reference value vdac _k of a comparator, an integrator accumulated value sum_k and a code stream accumulated value Bcode.

Calculating an output value ydiff_k of the comparator:;

Calculating an output value sum_k of the integrator: ;

judging the sign of the output value of the integrator, and accumulating the code stream:

;

Updating the reference value of the comparator:

;

the counter k is incremented by 1 and returns to step S222, and exits when k=n.

The step S3 specifically comprises the following steps of calculating a denoising result from the coding value Bcode:

;

Wherein yft is the denoising result.

Based on the three steps (steps S1-S3), denoising and filtering on the coordinate (i, j) are completed, each coordinate position on the infrared image Ysrc is traversed, and filtering on each point is completed.

It should be noted that:

for pixels at the boundary of the infrared image Ysrc, for example, when i < r, the ordinate of the top left corner vertex of the neighborhood is 1, the ordinate of the bottom right corner vertex of the neighborhood is D, and when i > H-r, the ordinate of the top left corner vertex of the neighborhood is H-D+1, and the ordinate of the bottom right corner vertex of the neighborhood is H.

For the pixels at the boundary of the infrared image Ysrc, for example, when j < r, the abscissa of the top left corner vertex of the neighborhood is 1, the abscissa of the bottom right corner vertex of the neighborhood is D, and when i > W-r, the abscissa of the top left corner vertex of the neighborhood is W-D+1, and the abscissa of the bottom right corner vertex of the neighborhood is W.

The neighborhood of the point is a space domain range, and can be expanded to a time domain range and a time-space domain range.

The sgm-delta quantization code mentioned in the invention is first order, and can be extended to higher order.

It should be noted that, for simplicity of description, the foregoing embodiments are all described as a series of acts, but it should be understood by those skilled in the art that the present application is not limited by the order of acts, as some steps may be performed in other orders or concurrently in accordance with the present application. Further, it should be understood by those skilled in the art that the embodiments described in the specification are preferred embodiments and that the actions involved are not necessarily required for the present application.

In the above embodiments, the basic principle and main features of the present invention and advantages of the present invention are described. It will be appreciated by persons skilled in the art that the present invention is not limited by the foregoing embodiments, but rather is shown and described in what is considered to be illustrative of the principles of the invention, and that modifications and changes can be made by those skilled in the art without departing from the spirit and scope of the invention, and therefore, is within the scope of the appended claims.

Claims (3)

1. The sgm-delta-based infrared image denoising method is characterized by comprising the following steps of:

S1, determining a point set which can participate in filtering in a neighborhood;

s2, carrying out sgm-delta quantization coding on the point set, wherein the step S2 comprises the following substeps:

S21, calculating a filter reference value Vref+, vref-, wherein the step S21 comprises the following sub-steps:

s211, calculating a mean value ymean of the point set { Y _ij }:

wherein n is the number of elements in the point set { Y _ij };

S212, calculating a weight average value Vref:

wherein w represents a weight inversely proportional to the neighborhood variance var, where the local variance is large is achieved, vref tends to the image value at the (i, j) point Where the local variance is small, vref tends to the neighborhood mean;

The relation between the weight and the neighborhood variance is:

Wherein eps is a constant set by the user;

the calculation formula of the local variance is:

;

S213, calculating a filtering reference value:

;

Wherein vd represents Is a fluctuation level of (2);

s22, traversing the point set { Y _ij }, calculating a coded value Bcode, wherein the step S22 comprises the following sub-steps:

S221, initializing variable values;

S222, calculating an output value of the comparator;

s223, calculating an output value of the integrator;

s224, judging the sign of the output value of the integrator and accumulating the code stream;

s225, updating a reference value of the comparator;

S226, returning to the step S222 until each element in the point set is traversed, wherein the initialization variable value comprises one or more of a counter k, a negative terminal reference value vdac _k of a comparator, an integrator accumulated value sum_k and a code stream accumulated value Bcode;

calculating an output value ydiff_k of the comparator: ;

Calculating an output value sum_k of the integrator: ;

judging the sign of the output value of the integrator, and accumulating the code stream:

;

Updating the reference value of the comparator:

;

the counter k accumulates 1 and returns to the step S222, and when k=n, the process is exited;

s3, calculating a denoising result through the coding value, wherein the step S3 specifically comprises the following steps of calculating the denoising result through the coding value Bcode:

;

Wherein yft is the denoising result.

2. The method for denoising an infrared image based on sgm-delta as set forth in claim 1, wherein step S1 is specifically to judge whether another point coordinate (i ', j') in the neighborhood of the coordinate (i, j) point satisfies a preset condition, and if so, to classify the value of the point (i ', j') into a point set { Y _ij };

wherein, the preset conditions are:

;

In the formula, Representing image values at coordinates (i, j) on infrared image Ysrc; Representing image values at coordinates (i ', j') on the infrared image Ysrc; Representing a threshold value, related to the noise level.

3. The method for denoising an infrared image based on sgm-delta according to any one of claims 1-2, wherein the neighborhood comprises a spatial domain, or a temporal domain, or a time-spatial domain.