JPH08202870A - Picture processing method - Google Patents

Abstract

PURPOSE: To obtain a noise reduction picture while minimizing the blur of an edge and density irregularity to a picture of low S/N. CONSTITUTION: This method executes a step 101 giving desired low pass filter processing to picture data, a step 102 calculating the magnitude of density change in the direction of straight line components concerning each straight line component passing through the picture element by each picture element of a low-pass-filter-processed picture, a step 103 obtaining the straight line component minimizing density change and a step 104 executing one-dimensional nonlinear smoothing processing by limiting to the straight line component of an original picture provided with the same direction as the straight line component of minimum density change obtained from the low-pass-filter-processes picture by each picture element of the original picture. Thus, as the straight line component of minimum density change in the low-pass-filter-processed picture is detected, a straight line component which is not affected by noise can be detected to one-dimensionally smooth by limiting to the straight line component provided with the directions. Thereby, while saving a false picture made by the structuring of noise components and minimizing the deterioration of space resolution, noise is effectively reduced.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to noise reduction processing in the field of image processing, and more particularly to a processing method for reducing noise without causing image blurring and density unevenness.

[0002]

2. Description of the Related Art Noise reduction processing basically has the function of smoothing abrupt changes in density, and therefore has the effect of blurring the outline of a figure. Therefore, there is a noise reduction process called edge-preserving smoothing, which roughly estimates the contour line of a figure by some method and performs smoothing in a manner that does not impair it. Various methods have been conventionally devised for this type of noise reduction processing. A typical example is the LLSE method (JS. Lee, Digital Image Enhancement and Noise Fil.
tering by Use of Local Satistics, IEEE, Trans on P
attern Anal.Machine Intell., Vol.PAMI-2 No.2, Marc
h, 1980).

This method defines the degree of presence or absence of an edge structure in a local area based on the local variance value, and adjusts the smoothing effect according to the degree of the edge structure. Specifically, for each pixel of the image I (i, j), the variance value σ _I (i, j) of the local region (25 pixels of the 5 * 5 matrix)
Is calculated and the degree of smoothing is adjusted according to the following formula. The following σ is a threshold value representing the noise level.

I ~ (i, j) = c (i, j) (I (i, j) -m _I (i, j)) + m _I (i, j) c (i, j) = σ _I (i, j) ² / (σ _I (i, j) ² + σ ² ) m _I (i, j) = (Σ ² _{k, l = -2} I (i + k, j + l)) / 5 σ _I (i, j) = (Σ ² _{k, l = -2} (I (i + k, j + l) -m _I (i, j)) ² ) / 5 This equation gives the variance value σ _I ( If i, j) is sufficiently larger than σ, then I ~ (i, j) ≈ I (i, j), and the variance value σ _I (i, j) is sufficiently smaller than σ. In this case, a value between the original pixel value and the simple mean value is determined according to the variance value σ _I (i, j), such as I ~ (i, j) ≈ m _I (i, j). To be done.

A method using local template matching is a method for smoothing after considering the local structure of an image with a little more precision (Junichiro Toriwaki "Digital image processing for image understanding [Ι]). Shokodo, November 30, 1989, pp112-114).

In this method, a typical pattern of edges and lines in a local area of an image is prepared in a template, and the neighborhood U ((i, j) of each pixel (i, j) of the image is prepared. ) Template matching is performed with the input density value of
This is a method of performing smoothing processing using a template adapted to the local structure of ((i, j)) as a weighting function. To explain in more detail, first, some partial image samples (templates) are prepared in advance. On the other hand, the input density values of the neighborhood U ((i, j)) of the pixel (i, j) arranged in a fixed order are regarded as a one-dimensional vector and are represented as F _ij . It is assumed that the template is also one-dimensionally vectorized in the same order, and it is A ₁ , A ₂ , ..., _Am .
At this time, k ₀ = minS (F _ij , A _k ) is calculated for the function S representing the goodness of fit of each template in (i, j), and the template A _k0 is used as a weighting function for the pixel. Smoothing processing is performed from the pixel values included in the neighborhood U ((i, j)) of (i, j). Here, a variety of concrete forms of the function S representing the goodness of fit are devised especially in the field of pattern recognition and numerical classification in statistics.

[0007]

DISCLOSURE OF THE INVENTION Problems to be Solved by the Invention
In the matching method, if the template size is small, it is not possible to draw out a global structure due to the local structure, and there is a risk that a false structure due to noise will be emphasized and structured and visualized. On the contrary, there is a problem that the image is blurred when the size of the template is increased. Further, in the LLSE method, since the structure of the image is captured by the local variance value, only the edge is determined and the morphological information is not taken into consideration. For this reason, in order to bring out the global structure, if the area for calculating the variance value is made large, the image tends to be blurred. Conversely, if the area for calculating the variance value is made small, it becomes sensitive to fluctuations due to noise and the noise reduction effect is small. There was a problem of becoming. The purpose of this patent is
It is to reduce the noise by preserving the detailed structure of the image as much as possible while making the global structure of the image stand out and be sharp against the image data buried in the noise.

[0008]

A step of subjecting image data to a desired low-pass filter processing, and for each pixel of the image subjected to the low-pass filter processing, for each linear component passing through the pixel, the linear component direction The step of calculating the magnitude of the density change of, and obtaining the straight line component that minimizes the magnitude of the density change, and the magnitude of the density change obtained from the low-pass filtered image for each pixel of the original image is the minimum. The step of performing the one-dimensional nonlinear smoothing process is limited to the straight line component of the original image having the same direction as the straight line component.

[0009]

In the low-pass filter image, the detailed structure and the noise component disappear and a global structure appears. Therefore, if the straight line component with the minimum density change is obtained in the low-pass filter image, the direction of the global iso-concentration line without the influence of noise can be captured. Therefore, if the original image is subjected to a one-dimensional non-linear smoothing process with a straight line component in the same direction as the straight line component obtained for each pixel of the low-pass filter image, it is smoothed in the global straight line direction without the influence of noise. Therefore, it becomes a smoothing in the direction that it should have, without being attracted to the noise component, and the global structure is exposed without structuring and visualizing the noise.
Also, in the clear edge structure part, even if a low-pass filter is loosely applied, the traveling line direction of the edge does not change,
One-dimensional smoothing in that direction does not destroy the edge.

[0010]

EXAMPLE An example of the present invention will be described with reference to FIG.
The target two-dimensional image is represented by I (i, j). In addition, the 5 * 5 matrix in FIG. 2 represents a linear component,
In the following examples, the change in density is examined for the linear component expressed by this matrix.

[Step 101] The low-pass filter L is applied to the two-dimensional image I (i, j), and the result is IL (i, j).
Here, the low-pass filter L is a convolution by the 3 * 3 matrix M shown in Formula 1.

[0012]

[Equation 1]

[Step 102] For each pixel (i, j) of the low-pass filtered image IL, the magnitude E (i) (i = 1 to 8) of the density change is calculated for the linear components in the eight directions shown in FIG. )
Is calculated according to the following equation 2.

Δ _{k, l} (i, j) = | IL (i + k, j + l) -IL (i, j) | (k, l = −2 to 2) ... (Equation 2) Direction 201: E (1) = Δ _0,1 (i, j) + Δ _{0, -1} (i, j) + Δ _0,2 (i,
j) + Δ _{0, -2} (i, j) direction 202: E (2) = Δ _-1,1 (i, j) + Δ _{+ 1, -1} (i, j) + Δ
_{+ 1, -2} (i, j) + Δ _{-1, + 2} (i, j) direction 203: E (3) = Δ _-1,1 (i, j) + Δ _{+ 1, -1} (i, j) + Δ
_{-2, + 2} (i, j) + Δ _{+ 2, -2} (i, j) direction 204: E (4) = Δ _-1,1 (i, j) + Δ _{+ 1, -1} (i, j) + Δ
_{+ 2, -1} (i, j) + Δ _{-2, + 1} (i, j) direction 205: E (5) = Δ _1,0 (i, j) + Δ _-1,0 (i, j) + Δ _2,0 (i,
j) + Δ _-2,0 (i, j) direction 206: E (6) = Δ _{-1, -1} (i, j) + Δ _{+ 1, + 1} (i, j) + Δ
_{+ 2, + 1} (i, j) + Δ _{-2, -1} (i, j) direction 207: E (7) = Δ _{-1, -1} (i, j) + Δ _{+ 1, + 1} (i , j) + Δ
_{+ 2, + 2} (i, j) + Δ _{-2, -2} (i, j) direction 208: E (8) = Δ _{-1, -1} (i, j) + Δ _{+ 1, + 1} (i , j) + Δ
_{+ 1, + 2} (i, j) + Δ _{-2, -1} (i, j) [Step 103] Represents a straight line component that minimizes the magnitude E (i) of density change (i = 1 to 8) The density value of 5 pixels is extracted from the original image I and is set as p (i) (i = 1 to 5). For example, if E (1) is the minimum, the following equation 3 is obtained.

P (1) = I (i, j + 2), p (2) = I (i, j + 1), p (3) = I (i, j), p (4) = I ( i, j-1), p (5) = I (i, j-2) (Equation 3) [Step 104] 5 pixel density value p (i) (i = 1 determined in Step 103) The smoothing process is performed by the LLSE method from (5) to (5). If this processing is expressed by a mathematical expression, it becomes (Equation 4).

I ~ (i, j) = c (i, j) (I (i, j) -m _I (i, j)) + m _I (i, j) c (i, j) = σ _I (i, j) ² / (σ _I (i, j) ² + σ ² ) m _I (i, j) = (Σ ⁵ _{k = 1} p (k)) / 5 σ _I (i, j) = ( Σ ⁵ _{k = 1} (p (k) -m _I (i, j)) ² ) / 5 (Equation 4) [Step 105] I to (i, j) are output images.

When the extraction of the direction component having the smallest density change in steps 102 to 103 is performed on the original image, the detection directions are different due to the influence of noise, and the original global direction is obtained. There are things that are not available. Therefore, even if one-dimensional smoothing is performed in the disparate detection directions, not only the noise reduction effect is weak, but in the worst case, noise may be intentionally structured. In this embodiment, first, the original image is subjected to a simple averaging process using a 3 * 3 matrix to reduce noise. In order to find the straight line direction in which the density change is the minimum from this noise-reduced image, FIG.
A direction component that is not affected by noise as in (b) can be extracted, and the smoothing direction is aligned with the original direction of the image. This makes it possible to reduce noise without structuring the noise. Here, step 101 of the present embodiment
By changing the characteristics of the low-pass filter L, the image quality of the processing result can be changed. In order to raise the global structure, the low-pass filter should be strengthened and the high-frequency component should be cut as much as possible. Conversely, if the detailed structure should be left, the low-pass filter should be weakened and the high-frequency component should be taken in as much as possible.

For example, the 3 * 3 matrix M of step 101 is

[0019]

(Equation 5)

Then, the low-pass effect is weakened,

[0021]

(Equation 6)

If so, the low-pass effect becomes weaker and the quality of the processed image approaches that of the original image. Further, in the present embodiment, the low-pass filter processing is performed by performing the convolution processing on the image itself, but the filter processing may be performed so as to pass a desired band in the frequency space.

Further, in the above step 105, the value of step 104 is used as it is as the output image, but in addition to that, the value of step 104 is mainly included, and other values are taken into consideration in the output image. You can also For example, the final output may be an image containing 20% of the low-pass image value of step 101 and 80% of the image obtained in step 104. Alternatively, the image obtained in step 104 may be low-pass filtered to be the final output.

[0024]

According to the present invention, it is possible to reduce the noise by preserving the detailed structure of the image as much as possible while making the global structure of the image stand out and clear for the image data buried in the noise. You can

[Brief description of drawings]

FIG. 1 is a flow chart showing a processing procedure of an embodiment of the present invention.
It is

FIG. 2 is a diagram showing a linear component in the embodiment of the present invention.

FIG. 3 is a diagram for explaining the effect of the image local structure direction detecting means of the present invention.

[Explanation of symbols]

101: A step of performing low-pass filter processing on the original image. 102: A step of calculating the magnitude of the density change of each linear component from the low-pass filtered image for each pixel. 103: A step of detecting a linear component having the smallest change in density. 104: A step of performing one-dimensional non-linear smoothing processing on the original image by limiting to the linear component in the same direction as the linear component detected in step 103. 201-208: One of the directional components.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Junichi Taguchi 1099 Ozenji, Aso-ku, Kawasaki-shi, Kanagawa Incorporated company Hitachi, Ltd. Systems Development Laboratory

Claims (7)

[Claims] 1. Image processing in smoothing image processing
The desired low-pass filter processing to the data, and for each pixel of the low-pass filtered image, for each linear component passing through that pixel, the magnitude of the density change in the direction of the linear component is calculated, and the density change Of a straight line component of the original image having the same direction as the straight line component having the smallest magnitude of the density change obtained from the low-pass filtered image, for each pixel of the original image. An image processing method including a step of performing one-dimensional non-linear smoothing processing limited to components. 2. In the step of performing the one-dimensional non-linear smoothing process, from the pixels on the straight line component of the original image having the same direction as the straight line component having the minimum density change obtained from the low-pass filtered image, The image processing according to claim 1, wherein the average value and the magnitude of the density change are obtained, and the output value of the pixel of interest is determined based on a desired function having the average value and the magnitude of the density change as variables. Method. 3. The desired function is F (q, where I is the pixel value of interest, and m and q are the average value and the magnitude of density change by the pixels on the straight line where the magnitude of the density change is minimum. The image processing method according to claim 1, which is a function F defined by) = c (Im) + mc = q ² / (q ² + σ ² ) (σ: constant). 4. The image processing method according to claim 1, wherein in the step of calculating the magnitude of density change in claim 1, E (i) defined by the following equation is used as the magnitude of density change. E (i) = Σ _k | p _i (0) -p _i (k) | where p _i (k) is the pixel value on the linear component that passes through the pixel of interest and evaluates the magnitude of the density change, k Is an ordered value of the range for evaluating the magnitude of change in density, such as 5 on a straight line.
When evaluating a point, k is five values of -2, -1, 0, 1, and 2, p _i (0) is the pixel value of interest, and p _i (-1) and p _i (1) are Pixel values on both sides of the pixel, p _i (-2) and p _i (2), are pixel values on the two sides next to the pixel of interest. 5. The image processing method according to claim 1, wherein the low-pass filter processing is performed by subjecting image data to convolution with a matrix of a predetermined size. 6. The matrix of the predetermined size is 3 * 3.
The image processing method according to claim 5, wherein the image is a square matrix. 7. In the low-pass filter processing, frequency data obtained by Fourier transforming image data is multiplied by a function of a desired pass characteristic, and then the resulting data is subjected to inverse Fourier transform. The image processing method according to claim 1, wherein conversion is performed.