JPH1185961A - Image processor and image processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To increase effects of contrast emphasis and noise reduction at the same time by finding the presence probability of a linear shade by pixels according to the results of two kinds of spatial image process and changing characteristics of temporal image processes by the pixels according to the presence probability. SOLUTION: For example, the breast part 4a of a body 4 to be inspected which is mounted on a bed 5 is irradiated with X rays 3 radiated by an X-ray source 2. Transmitted X rays 3a which are shaded owing to difference in X-ray absorptivity among respective parts while transmitted through the breast part 4a are introduced into an image intensifier 6 and a converted and amplified optical image is picked up by a TV camera 7. For a digital X-ray image signal 8, an image processor 9 performs two kinds of image process by which the effect of emphasis of a linear shade is recognized basically for the same digital X-ray image. Totally judging from the result, the presence probability of the linear shade is found by the pixels and according to the presence probability, characteristics of a time file are changed by the pixels.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to an image processing apparatus and method for processing, for example, a digital X-ray image including a linear shadow such as a blood vessel or a guide wire.

[0002]

2. Description of the Related Art As a medical technique using X-rays, catheter treatment under fluoroscopy is actively performed. However, X-ray fluoroscopy reduces the dose to patients as compared with radiography. , Noise superimposed on the image is large, and linear shadows such as catheters, guidewires and blood vessels,
There was a problem that it was difficult to see because of the background noise.

Further, when fluoroscopy is performed with an increased dose, background noise is relatively small, and visibility of linear shadows is high. However, there is a drawback that the exposure of patients and medical technicians increases. . For this reason, techniques for reducing noise and improving contrast by image processing have been studied even now.

The most commonly used image processing technique for noise reduction is a method of averaging a plurality of temporally continuous images. This technique is already known, and a regression filter called a recursive filter is widely and generally used as an applied technique.

However, such a temporal filter can effectively reduce only noise for a subject with little movement, but for a relatively moving subject such as a guide wire inserted into a cardiovascular vessel. However, there are disadvantages in that the contrast is reduced or an afterimage is generated. For this reason, techniques have been developed that detect the movement of the subject and change the filter characteristics where there is movement (Japanese Patent No. 250878, Japanese Patent Publication No. 6-69447, Japanese Patent Application Laid-Open No. 3-198836, JP-A-6-473
No. 05, JP-A-7-79956, JP-A-8-
255238).

However, in these techniques, since the movement of the subject is detected based on the difference between the current image and the past image, a change in pixel value due to noise is also detected as the movement of the subject. There was a problem. In particular, when using a small-diameter catheter or a low-contrast guide wire, if the motion detection accuracy is increased,
There has been a problem that noise reduction cannot be performed effectively.

On the other hand, a spatial filter is generally used for the purpose of reducing noise and improving contrast. This spatial filter can be used in various ways, for example, by taking a weighted average of pixels in the vicinity of the local area to reduce and smooth noise, or to enhance contrast by spatial differentiation. However, smoothing blurs not only the noise but also the linear shadows such as guidewires and catheters that are important, and the contrast is impaired.In the case of contrast enhancement, the noise is also enhanced. There is a disadvantage that it will.

For this reason, an apparatus for spatially recognizing a blood vessel shadow pattern (Japanese Patent Laid-Open No. 4-122355), or an apparatus for detecting an edge area of a subject and applying contrast enhancement to the edge area and smoothing the other area. (JP-A-60-245084) and the like have been developed.

[0009] However, the former method of pattern recognition has the following problems.
Since many averaging circuits and comparison circuits are required for each pixel, a very large number of circuits are required for performing real-time operations such as fluoroscopy processing, and furthermore, there is no effect of reducing noise.

In the latter method of edge enhancement, only one method is used.
In order to distinguish the edge area from the background area based on the bit information, if there is a misrecognition, the misrecognized noise is output adjacent to the extremely smoothed part with contrast enhancement, etc. And misrecognition is sensitive to artifacts in the output image, which tends to impair the image quality.

[0011]

As described above, in the prior art, the effects of the afterimage reduction and the noise reduction are simultaneously enhanced for the image of the moving subject, and the effects of the contrast enhancement and the noise reduction are improved. At the same time it was difficult to raise.

The present invention realizes simultaneously enhancing the effects of afterimage reduction and noise reduction, and simultaneously enhances the effects of contrast enhancement and noise reduction, and provides guidewires, catheters, blood vessel images, etc. in X-ray medicine. It is an object of the present invention to provide an image processing apparatus and method capable of favorably displaying a shadow of an image.

[0013]

According to the present invention, an image containing linear shadows and noise is subjected to at least two types of spatial image processing, and a result of the two types of spatial image processing is obtained. , The existence probability of a linear shadow is obtained for each pixel, and the characteristic of temporal image processing is changed for each pixel according to the existence probability.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An image processing apparatus according to the present invention will be described below with reference to the drawings. Here, as image processing targets, X including linear shadows such as blood vessels, guide wires, and catheters and background noise is used.
The explanation will be given using a fluoroscopic image as an example.

FIG. 1 shows the configuration of an X-ray fluoroscope incorporating the image processing apparatus according to this embodiment. The X-ray fluoroscope main body 1 irradiates the X-rays 3 emitted from the X-ray source 2 toward, for example, the chest 4a of the subject 4 placed on the bed 5 and transmits the X-rays 3 through the chest 4a. Then, the transmitted X-ray 3a shaded by the difference in the X-ray absorptance of each part is introduced into the image intensifier 6, and the converted and amplified optical image is captured by the TV camera 7. I have. The output signal of the TV camera 7 passes through a preamplifier, an analog / digital
It is converted into a digital X-ray image signal 8.

The digital X-ray image signal 8
Is appropriately processed by the image processing device 9, and by this processing, the background noise is reduced, and the linear shadows are relatively emphasized, and thereafter, are displayed in shades on the CRT display 10.

By the way, blood vessels and other catheters and guide wires to be inserted into the body are represented on a digital X-ray image by thin linear shadows according to their outer shapes. As described in the prior art, such a thin linear shadow is
Its properties, for example, its spatial frequency and temporal frequency are as high as noise, so it is susceptible to noise reduction processing, and it disappears, blurs, or loses its contrast with noise. There's a problem. Conventionally, such a problem has been addressed by increasing the X-ray irradiation dose and improving the signal-to-noise ratio (S / N).

On the other hand, in the present embodiment, even if the X-ray irradiation dose is not increased or further reduced, the contrast of the linear shadow is not impaired, and only the background noise is effectively reduced. Is what you do. Thus, if background noise is reduced and linear shadows can be observed with sufficient contrast, it is expected that the accuracy of medical treatments such as catheterization and biopsy performed under this fluoroscopy will be significantly improved. .

In the present embodiment, basically, the same digital X-ray image is subjected to at least two types of image processing which has an effect of enhancing a linear shadow, and two types of image processing based on these two types of image processing are performed. Judging comprehensively from the result, the existence probability of the linear shadow is obtained for each pixel, and the characteristic of the time filter is to be changed for each pixel according to this existence probability. For pixels with a relatively high pixel density, the smoothing effect of the temporal filter is kept low to prevent a decrease in the contrast of linear shadows.On the other hand, the probability of linear shadows being relatively low and the probability of being background noise being relatively low For high pixels, the smoothing effect of the temporal filter is enhanced to effectively reduce noise.

FIG. 2 shows a processing procedure of the image processing apparatus 9. This processing may be realized by software using a program written in a computer-readable storage medium, or may be realized by hardware by combining circuits corresponding to the respective blocks in FIG. 2 according to the processing procedure. You may make it. The image processing device 9
It is more preferable to configure so that an input image can be processed and displayed in real time.

As an input image, a digital X-ray image I ₀ is continuously supplied from the TV camera 7 to the image processing device 9 at a predetermined cycle. First, from the digital X-ray image I ₀ , the existence probability of a linear shadow is obtained for each pixel (a unity of the existence probabilities of all pixels is
"Existence probability distribution"). In FIG. 2, this existence probability distribution is indicated as “F3”. In this embodiment,
In order to obtain the existence probability distribution F3, two types of image processing of the A sequence and the B sequence are employed. First, the processing procedure for the A-series will be described.

The sequence A is A1, A2, A3,
It consists of five steps A4 and A5. First, in step A1, the digital X-ray image I _0, subjected to a treatment that minimum value filtering, to obtain the minimum value image I _min. As is well known, as shown in FIG. 3, the minimum value filtering is performed by moving a filter mask of, for example, a 2 × 2 matrix so as not to overlap, and at each position, 4
In this process, the smallest pixel value is selected from the two neighboring pixels, and the smallest value is used as the pixel value of the corresponding pixel in the minimum value image _Imin .

[0023] By this process, to reduce the matrix size to 1/4, for example, when the matrix size of the digital X-ray image I ₀ is the size of 1024 × 1024 is reduced to a matrix size of 512 × 512 However, the size of the circuit can be reduced by performing the subsequent processing with a matrix size of 512 × 512. According to such minimum value filtering, the contrast of the linear shadow is slightly enhanced.

Next, in step A2, a high-pass type (high-pass type) spatial filtering is performed on the minimum value image I _min obtained in A1, and the minimum value obtained by filtering a component having a relatively low spatial frequency is filtered. Obtain the image I _high .

The minimum value image I _high subjected to the high-pass filtering in step A2 and the minimum value image I _min obtained in step A1 are subjected to a linear operation represented by the following equation (1) in step A3. In other words, the difference image I _sub1 is obtained by executing the weighted subtraction (the difference is obtained by multiplying the weight).

I _sub1 = W ₁ × I _min −W ₂ × I _high + C (1) FIG. 4A shows a high-pass filtered minimum value image I
FIG. 4B shows a spatial profile (position function of pixel value) of each of the _high value and the minimum value image I _min , and FIG. 4B shows the spatial profile of the difference image I _sub1 obtained by the difference processing of the above equation (1) Profile. By appropriately setting the coefficients W ₁ , W ₂ , and C in the above equation (1), FIG.
As shown in (b), on the difference image I _sub1 , it is possible to invert the polarity of only the portion estimated to have a relatively high probability of being a linear shadow with respect to the other portions.

The difference image I _{sub1 in} which the part which is estimated to have a relatively high probability of being a linear shadow and the other part are identified by the polarity is binarized according to the polarity.
A portion which is estimated to have a relatively high probability of being a linear shadow is represented by “1”, and the other portions are represented by “0”.
A value image I _bi1 is obtained (step A4).

Since the binary image I _bi1 contains background noise in most cases, it is necessary to remove the background noise. Here, as shown in FIG. 5A, the noise is isolated, while the linear shadow has high connectivity.
Utilizing such a difference in properties, here, pattern matching is performed on the binary image I _bi1 (step A5). As is well known about pattern matching,
A plurality of types of linear shading patterns as shown in FIGS. 5B to 5E are prepared in advance, and a portion that matches this pattern is left as it is, and a pixel value that does not match is set to “0”.
This is a process of inverting to. By such pattern matching, as shown in FIG. 5 (f), most of the noise is removed, and only the portion where the probability of the presence of the linear shadow is estimated to be relatively high can be extracted. The binary image obtained in the above-described A series is referred to as “F1”.

Next, the B series will be described. The sequence B is, like the sequence A, processed in order of B1, B2, B3, B
4 and B5. First, step B
In step 1, the digital X-ray image I ₀ is subjected to a process called spatial smoothing filtering to obtain an average value image I 0.
_{Get ave} . As shown in FIG. 6, for example, the smoothing filtering is performed by moving a filter mask of, for example, a 2 × 2 matrix so as not to overlap, and at each position, an average value of pixel values of four neighboring pixels in the mask Is calculated, and the average value is used as the pixel value of the corresponding pixel in the average image _Iave .

[0030] By this process, to reduce the matrix size to 1/4, for example, when the matrix size of the digital X-ray image I ₀ is the size of 1024 × 1024 is reduced to a matrix size of 512 × 512 However, the size of the circuit can be reduced by performing the subsequent processing with a matrix size of 512 × 512. According to such smoothing filtering, the background noise is reduced, and accordingly, the linear shadow is easily recognized.

Next, in step B2, a low-pass type (low-pass type) spatial filtering is performed on the average image I _ave obtained in B1, and the average value obtained by filtering a component having a relatively low spatial frequency is filtered. Obtain the image I _low .

The average value image I _low subjected to the low-pass filter in step B2 and the average value image I _ave obtained in step B1 are subjected to a linear operation represented by the following equation (2) in step B3. In other words, by executing the weighted subtraction (multiplying the weight to obtain the difference), the difference image I _sub2 is obtained.

I _sub2 = W ₃ × I _ave −W ₄ × I _low + D (2) FIG. 7A shows a low-pass filtered average value image I
and _low, demonstrating spatial profile of the respective average image I _ave, in FIG. (b), shows the spatial profile of the difference image I _sub2 obtained by differential processing of the above formula (2). By appropriately setting the coefficients W ₃ , W ₄ , and D in the above equation (2), as shown in FIG. 7B, if the probability of a linear shadow on the difference image I _sub2 is relatively high. The polarity of only the estimated portion can be inverted with respect to the other portions. In the example of FIG. 7B,
It is estimated that four regions are relatively likely to be linear shadows. Of these regions, the second region from the left is a portion where a linear shadow is correctly identified, and the remaining three regions are: This is the part that was misidentified.

The difference image I _{sub2 in} which the part which is estimated to have a relatively high probability of being a linear shadow and the other part are identified by polarity is binarized according to the polarity.
A portion which is estimated to have a relatively high probability of being a linear shadow is represented by “1”, and the other portions are represented by “0”.
A value image I _bi2 is obtained (step B4).

In most cases, background noise also remains in the binary image I _bi2 , so that it is necessary to remove the background noise, and perform pattern matching on the binary image I _bi2 as in A5. (Step A5). Most of the noise is removed, and only the portion that is estimated to have a relatively high probability of the presence of the linear shadow can be extracted. The binary image obtained in the above-described B sequence is referred to as “F2”.

As described above, the two images obtained by the two systems of image processing are
In step AB, a quaternary flag image F3 is created from the binary images F1 and F2. The pixel values of the flag image F3 are assigned as shown in the following Table 1 according to a combination of the pixel value of one binary image F1 and the pixel value of the other binary image F2 with respect to the pixel at the same position. .

[0037]

[Table 1]

In both of the two types of binary images F1 and F2, a portion estimated to have a relatively high probability of being a linear shadow is represented by "1", and the other portions are represented by "0". Therefore, when the pixel value is “1” in both of the two types of binary images F1 and F2, the pixel is considered to have the highest probability of being a linear shadow. The pixel is given a pixel value of “3”.

Conversely, when the pixel value is "0" in both of the two types of binary images F1 and F2, the pixel is considered to have the lowest probability of being a linear shadow, and here, the flag image The pixel value of “0” is given to the corresponding pixel of F3.

In addition, since the minimum filtering is generally higher in the extraction ability of the linear shadow than the smoothing filtering, the binary image F1 through the minimum filtering is extracted.
In the binary image F2 that has been subjected to the smoothing filtering, if it is “0”, it is determined that the probability of being a linear shadow is the second highest. A pixel value of “2” is given to the pixel, while “0” is given to the binary image F1 subjected to the minimum value filtering, while “1” is given to the binary image F2 subjected to the smoothing filtering. In the case, the probability of being a linear shadow is 3
Here, the pixel value of “1” is given to the corresponding pixel of the flag image F3 as the second highest.

Next, a digital X-ray image I is applied to a time filter called a recursive filter (recursive filter).
₀ or an image derived therefrom by image processing is passed. In this embodiment, this recursive filter is
Two systems are provided, one for background noise and one for linear shadows (C
2, D2). As is well known, the recursive filter is a time filter of a type in which a previous filter result is reflected in the current filtering, and is defined by the following equation (3). Note that “a” represents the previous filter result,
This is a recursive coefficient that determines the filter characteristic of how much is reflected in the current filtering. Here, the lower the “a”, the more strongly the previous filtering result is reflected in the current filtering.

Y _i (x, y) = a × X _i (x, y) + (1−a) × Y _i−1 (x, y) (i = 1, 2,..., ∞) Y _i represents the ith recursive filtered image, and X _i is the ith digital X-ray image I _{0 of} the current frame supplied from the TV camera 7 to the image processing device 9 or derived therefrom by image processing. An image, for example, three images I ₀ , I _min , and I _ave are used to represent an inter-frame weighted average image given by the following equation (4). Note that I _min and I _ave are used in equation (4) after the number of pixels is converted into a matrix size equal to I ₀ . Further, the coefficients α, β, and γ may be fixed coefficients or may be configured to be changed for each pixel according to the value of the flag image F3.

X _i (x, y) = α × I ₀ (x, y) + β × I _min (x, y) + γ × I _ave (x, y) (4) (α + β + γ = 1) Is changed for each pixel according to the pixel value of the flag image F3, and the way of changing is different between the recursive filter for background and the recursive filter for linear shadow. This is due to the difference in purpose between the two filters. Since the background recursive filter aims to reduce the background noise, the recursive coefficient a is assigned from a relatively low range and the previous filter result is obtained. More strongly reflected (C
2) Conversely, the recursive filter for linear shading aims to enhance the linear shading, and if the result of the previous filter is too strongly reflected, “blur” due to the movement at the time of insertion of a catheter or the like may occur. Therefore, the recursive coefficient a is set relatively high so that the previous filter result is not reflected too strongly (D
2).

More specifically, as shown in Table 1 above, the recursive coefficient a of the background recursive filter
Is “0.3” when the flag is 3, “0.7” when the flag is 2, “0.6” when the flag is 1,
When the flag is 0, “0.5” is assigned. The background recursive filter is basically configured such that the higher the existence probability of the linear shadow is, the higher the recursive coefficient is and the less the effect of the afterimage is.
Output image I of recursive filter for linear shading at 1, E2
Portion value of ₁ pixel is output as the output image I ₃ of the output image forming circuit (in this embodiment, the portion of the flag is 3)
Is assigned a very small recursive coefficient,
It is configured to be a blurred image.

When the flag is 3, the recursive coefficient a of the linear shading recursive filter is set to “0.
9 ", all at other times" assign 0.8 ". Linear shadow for recursive filter, the value is the output image formation of pixel output image I ₁ of the linear shadow for recursive filtering in step E1, E2 below (in the present embodiment, the flag is 3 parts of) the portion to be output as the output image I ₃ of the circuit assigned a very large recursive coefficients for, is configured so that the influence of the afterimage is reduced.

[0046] By such recursive filtering process, two kinds of images of the image I ₁ of the background noise is greatly reduced, and the image I ₂ of the linear shadow is sufficiently emphasized blurred is suppressed due to the motion Is obtained. In this embodiment, the output of the recursive filter is used as the image I ₂ representing the linear shadow, but the original image I ₀ and the minimum value image I ₂ are used instead of the output image of the recursive filter.
_It is good also as a structure which uses any of _min and the average value image _Iave .

Next, at steps E1 and E2,
Either _{one of} the types of images I ₁ and I ₂ is selected for each pixel according to the flag indicating the existence probability of the linear shadow of the flag image F3. Specifically, as shown in Table 1, the flag is a 0, 1, 2 parts, selects the pixel value of the output image I ₁ of the recursive filter for the background as the pixel value of the image I _3, the flag is 3 the portions are configured so as to select the pixel value of the output image I ₂ of the recursive filter for linear shading as the pixel value of the image I _3. At this time, the output image I ₃ has a sufficiently reduced background noise and a clear image with a sufficient linear contrast.
Further, even when the catheter, the guide wire, and the like move in the image I ₀ , the image hardly contains the afterimages.

The spatial filter coefficient and filter size, the pattern matched filter and filter size, the recursive coefficient, and the like are the spatial thickness and density of the linear shadow to be recognized, the frequency characteristics of background noise, I. Optimal processing can be performed by appropriately determining the size of the field of view and the fluoroscopic X-ray conditions.

In general, the circuit scale of hardware is such that many averaging circuits and comparison circuits operate simultaneously,
When the generation of the pixel address, which is the position designation signal on the image, is complicated, it becomes large, and the possibility of practical real-time calculation becomes small. The pattern recognition as seen in the prior art is a configuration that is not suitable for real-time recognition. However, in the present embodiment, the pattern recognition of linear shadows is performed by comparing a linear spatial filter having the same number of bits of the accompanying information with a linear spatial filter located at the subsequent stage. And a binary filter applied to the binarized image, it is possible to perform processing by sequential scanning, and the circuit scale is small.

Further, in the present embodiment, the existence probability of the linear shadow is determined using the image processing of two forms, but the existence probability of the linear shadow is determined based on the selection result of the image processing of two or more systems. It may be configured.

The present invention can be implemented in various modifications without being limited to the above-described embodiment. For example, in the above description, an X-ray fluoroscopic image is targeted, but DSA
In many cases, it is not possible to form a recursive filter in a photographed image such as that described above or one-shot X-ray photographing. This is because images are not input continuously, and intermittent or only one image is input. In this case, a spatial filter can be used instead of a recursive filter. Image I ₁ is relatively intensely smoothed and I ₂ is a contrast-enhanced image, or the contrast of linear shadows is not enhanced or impaired using a minimum image, and the background is smoothed. Is done. Even in this case, the smoothing for the background is F1 =
Pixels with accompanying information of 0, F2 = 0 increase the smoothing effect, and F1 = 1, F2 = 0, or F1 = 0,
It is necessary to set the pixels of the incidental information of F2 = 1 so as not to be so large so as to suppress the contrast deterioration of the linear shadow that may be present in the background image. By forming an image based on such a design, it is possible to provide an observer with an image with reduced background roughness while maintaining the contrast of blood vessels and stomach walls.

[0052]

According to the present invention, at least two types of spatial image processing are performed on an image containing linear shadows and noise, and a line is obtained based on the result of the two types of spatial image processing. Since the probability distribution indicating the existence probability of the linear shadow is obtained, the presence or absence of the linear shadow can be determined with relatively high accuracy. By changing the characteristics of temporal image processing for each pixel or neighboring group according to this probability distribution, noise can be effectively reduced, and at the same time, linear shapes such as guide wires, catheters, and blood vessel shadows in X-ray medicine can be reduced. The contrast of the shadow can be effectively enhanced.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an X-ray fluoroscope incorporating an image processing apparatus according to an embodiment of the present invention.

FIG. 2 is a flowchart showing a processing procedure of the image processing apparatus according to the embodiment of the present invention.

FIG. 3 is a supplementary explanatory diagram of the minimum value filtering of FIG. 2;

FIG. 4 is a diagram showing profiles of a minimum image I _min of FIG. 2, a minimum image I _high subjected to a high-pass spatial filter, and a subtraction image I _{sub1 of} both of them.

FIG. 5 is a supplementary explanatory diagram of the pattern matching in FIG. 2;

FIG. 6 is a supplementary diagram of the median filtering of FIG. 2;

FIG. 7 is a diagram showing profiles of the average image I _ave of FIG. 2, an average image I _low subjected to a low-pass spatial filter, and a subtraction image I _{sub2 of} both of them.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... X-ray fluoroscope main body, 2 ... X-ray source, 3 ... X-ray, 3a ... Transmission X-ray, 4 ... Subject, 4a ... Chest, 5 ... Bed, 6 ... Image intensifier, 7 ... TV camera, 8 ... Digital X-ray image signal 9 ... Image processing device 10 ... CRT display

Claims (10)

[Claims] 1. A means for performing at least two types of spatial image processing on an image including a linear shadow and a noise, and based on a result of both the two types of spatial image processing. Means for calculating the existence probability of the linear shadow for each pixel. 2. A means for performing temporal image processing on the image or an image derived therefrom by image processing, and means for changing the characteristics of the temporal image processing for each pixel according to the existence probability of the linear shadow. The image processing apparatus according to claim 1, further comprising: 3. The spatial image processing means comprises a first image processing system and a second image processing system, wherein the first image processing system has a filter for enhancing the linear shadow. 2. The image processing apparatus according to claim 1, wherein the second image processing system includes a filter for reducing the noise. 4. The image processing system according to claim 1, wherein the first image processing system includes a high-pass filter that extracts a component having a relatively high spatial frequency from the image in which the linear shadow is enhanced, and a high-pass filter that extracts the high-pass component from the image in which the linear shadow is enhanced. A first difference circuit for subtracting the image processed by the filter, and a circuit for binarizing the image differed by the first difference circuit according to the polarity;
The second image processing system differentiates between a low-pass filter that extracts a component having a relatively low spatial frequency from the noise-reduced image and an image that has been processed by the low-pass filter from the noise-reduced image. 4. The image processing apparatus according to claim 3, further comprising a second difference circuit, and a circuit that binarizes the image obtained by the second difference circuit in accordance with a polarity. 5. The method according to claim 1, further comprising:
The image processing apparatus according to claim 3, further comprising a unit configured to select and output one of the image processing system and the second image processing system. 6. The means for changing the characteristic of temporal image processing, for a pixel having a relatively high existence probability, makes the smoothing characteristic of temporal image processing relatively low, 2. The image processing apparatus according to claim 1, wherein a smoothing characteristic of the temporal image processing is set relatively high for a pixel having a relatively high value. 7. Applying at least two types of spatial image processing to an image including linear shadows and noise, and based on a result of both the two types of spatial image processing. Obtaining the existence probability of the linear shadow for each pixel. 8. a step of performing temporal image processing on the image or an image derived from the image processing by the image processing; and changing a characteristic of the temporal image processing for each pixel or a neighboring group according to the probability distribution. The image processing method according to claim 7, further comprising: 9. A first selection means for selecting a linear shadow area and another area from an input image, and a linear area from the input image by a different kind of selection method from the first selection means. A second selection unit for selecting the input image; an existence probability determination unit for obtaining an existence probability of the linear shadow based on an output of the first selection unit and the second selection unit; and performing a noise reduction process on the input image. An image processing apparatus, comprising: noise reduction processing means for switching at least three types of filter coefficients based on the output of the existence probability determination means. 10. The image according to claim 9, wherein the noise reduction processing means changes a filter coefficient so that the noise reduction effect increases as the existence probability decreases in the input image. Processing equipment.