JP3485454B2 - Image gradation conversion device, image gradation changing method, medium recording program for executing the method, and infrared camera - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image gradation conversion device, an image gradation conversion method, and a medium recording a program for executing the method, and more particularly to a super multi-gradation image to a low gradation image. The present invention relates to a gradation conversion device for conversion, an image gradation conversion method, and a medium recording a program for executing the method.

[0002]

2. Description of the Related Art Infrared images taken by an infrared camera, CT
Some medical images captured by (Computed Tomography) and the like have a super multi-gradation of 4096 (12 bits), for example, in order to realize high resolution. However, the gradation that can be displayed on a general-purpose display that is generally used is 256 (8 bits) at most, so when displaying an image with super-multi-gradation on a general-purpose display, gradation conversion that reduces the gradation is performed. However, when the simple linear conversion is used, the following problems occur.

[0003]

For example, in an infrared image taken by an infrared camera, an image of an object having a relatively high temperature (for example, a human) has high brightness, but an object having a relatively low temperature ( For example, the background image has low brightness, so when simple linear conversion is used, an object with a high temperature has a high brightness image with little gradation change, and an object with a low temperature has a low brightness with little gradation change. It becomes an image, and detailed information (for example, gradation information indicating a smooth gradation change) is not clearly displayed.

[0004]

A first invention is a separating means.
A low frequency component representing the global gradation change an original image having a super multi-gradation manner, is separated into a high-frequency component representing the local gradation variations. At that time, the edge strength , which is the difference in gradation between the adjacent images of the original image, is detected by the edge strength means , and the separation characteristic is determined.
The changing means changes the separation characteristic for separating the original image into the low frequency component and the high frequency component according to the edge strength.

In the first aspect of the invention, loss of edge information from low frequency components is prevented by changing the separation characteristic according to the edge strength of the original image. The first invention is
Further, the gradation of the low frequency component separated by the separation step is reduced, and the high frequency component separated by the separation step and the low frequency component with the reduced gradation are combined.

[0006] In the first invention, it is possible to prevent the fine gradation information of the peripheral edge is lost if further displaying the super multi-tone image to a low tone display. According to a second aspect of the invention, in the separating step, the low frequency component extraction characteristic for extracting the low frequency component from the original image is changed in accordance with the edge strength detected in the edge strength detecting step, and the original image and the extracted low frequency component are changed. The high frequency component is extracted based on and.

According to the second aspect of the invention, information having a large gradation difference is incorporated in the low frequency component. According to a third aspect of the present invention, a pixel is divided into an edge peripheral pixel within a predetermined pixel number width centered on an edge and other non-edge peripheral pixels, and an edge preservation filter extracts a low frequency component from the edge peripheral pixel. Then, for the non-edge peripheral pixels, the low frequency component is extracted by the edge non-conservation filter.

According to the third aspect of the present invention, by separating the original image into edge peripheral pixels and non-edge peripheral pixels in advance, it is possible to determine whether to use the edge-preserving filter or the edge-preserving filter when extracting the low-frequency component. It is not necessary to make a judgment for each pixel. In the fourth invention, the size of the filter is reduced as the edge strength is higher. In the fourth aspect of the present invention, when the edge strength is high, a filter having a small size is used to extract the low frequency component while retaining the edge information, and when the edge strength is low, a filter having a large size is extracted. The low frequency component is extracted by using.

According to a fifth aspect of the present invention, a pixel is separated into an edge peripheral pixel having an edge as a center and a non-edge peripheral pixel other than the edge peripheral pixel. On the other hand, a low-frequency component is extracted by a filter having a large size. In the fifth aspect of the present invention, it is not necessary to determine the size of the filter used for extracting the low frequency component by separating the original image into edge peripheral pixels and non-edge peripheral pixels in advance.

A sixth aspect of the present invention captures a super multi-tone image,
After the low-frequency component and the high-frequency component are separated, the low-frequency component is converted into a smaller number of gradations and the high-frequency component is emphasized.
Then, the low frequency component and the high frequency component are combined and displayed on the low gradation monitor. In the sixth aspect , the super-multi-tone image is displayed on the low tone monitor without losing the fine luminance information.

The seventh aspect of the present invention is the invention of the first to fifth aspects.
A medium for recording a program for executing one of them is provided.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows an image of ultra-multi-gradation (M gradation) for solving the above-mentioned problems.
FIG. 9 is an explanatory diagram of a gradation conversion method for converting into an image of M),
Image pickup means 10 for picking up an original image Separating section 11 for separating an original image into a high frequency component representing a local brightness change of the original image and a low frequency component representing a global brightness change, and the gradation of the low frequency component is M Low frequency component gradation conversion unit 1 for converting from gradation to N gradations
2 (For the conversion method, linear conversion, non-linear conversion (γ correction method), Histogram Projection method, Histogram Equalizati
A well-known method such as an on method is applied), an emphasis processing unit 13 that performs emphasis processing (for example, linear conversion) on a high frequency component, a low frequency component converted to N gradations and a high frequency component that has been emphasized are added. After that, it is composed of a synthesizing / re-gradation converting section 14 for converting again to N gradations.

FIG. 1 also exemplifies a luminance change waveform along one horizontal scanning line of the original image. (A) shows a luminance change waveform (solid line) and a low frequency component waveform (broken line) of the original image. ) And (b) are low-frequency component waveforms after gradation conversion, (c) is a high-frequency component waveform after emphasis processing, (d) is a composite waveform of low-frequency components and high-frequency components, and (e) is re-gradation conversion. The following output waveform is shown.

First, by performing a smoothing operation on the original image information picked up by the image pickup means 10 using an average value filter, a Gaussian filter, etc., the maximum luminance is M1,
The low-frequency component waveform of the original image whose minimum brightness is M2 ((a)
The broken line) is obtained. In the low frequency component gradation conversion unit 12, this low frequency component waveform is converted into (M1−M2 + 1) steps of brightness into N steps of brightness and becomes a low frequency component (b) after gradation conversion.

On the other hand, the amplitude of the high-frequency component waveform calculated by subtracting the low-frequency component component waveform (dashed line in (a)) before gradation conversion from the original image information is doubled in the emphasis processing section 13, for example. Then, the high-frequency component waveform (c) is obtained after the emphasis processing. The synthesis / re-tone conversion section 14 adds the tone-converted low-frequency component (b) and the post-enhancement high-frequency component (c) to perform synthesis to generate a synthesized waveform (d), and the luminance becomes negative. The output waveform (e) is obtained by performing gradation conversion again so that the waveform after cutting is within N levels of luminance.

However, in the above method, when the noise removal processing which is indispensable for gradation conversion is executed on the original image having a portion where the luminance change is extremely large (hereinafter referred to as an edge), the fine luminance of the original image is around the edge. It is impossible to avoid loss of changes (gradation information). Figure 2
Is an explanatory diagram of the problem, and when a smoothing filter is applied to a portion where the luminance changes in M gradation steps (a ′) for noise removal processing, the step change does not occur in the low frequency component (b ′). A ramp-shaped change. When the size of the filter is increased to ensure noise removal, the slope of the ramp portion becomes gentle.

The high frequency component (c ') is {(a')-
Since it is obtained as (b ′)}, the waveform is such that it gradually increases from zero in the positive direction, then reverses stepwise to the negative side, and then gradually returns to zero. Therefore, the gradation-converted low-frequency component (b ″) obtained by performing gradation conversion on the low-frequency component (b ′) and the post-enhancement-process high-frequency component (c ″) obtained by performing emphasis operation on the high-frequency component (c ′) In the combined waveform (d ') obtained by combining and, overshoot and undershoot indicated by diagonal lines occur.

When the negative waveform (that is, the undershoot portion) of the composite waveform (d ') is cut and then the output waveform (e') obtained by performing the gradation conversion again is displayed on the general-purpose display, the (a) portion is displayed. Is displayed with higher brightness (that is, brighter) than the original image, and the part (b) is displayed with the lowest brightness (that is, black). That is, it is unavoidable that a high-intensity band is generated on one side of the edge and a minimum-intensity band is generated on the other side of the edge, and fine luminance change information around the edge is lost. This is because a noise removal filter having a constant size is applied to the entire original image when obtaining (that is, when removing noise from the original image).

The present invention has been made in view of the above problems, and an image gradation in which a fine luminance change around an edge is not lost when converting a super-multi-gradation image into an image of normal gradation. An object of the present invention is to provide a conversion device, an image gradation conversion method, and a program recording medium recording a program for executing the image gradation conversion method. FIG. 3A is a configuration diagram of one embodiment of the gradation conversion device according to the present invention,
The infrared camera 31, the microcomputer 32, the display 33 capable of displaying a relatively small number of luminance levels (for example, 256) and the external program recording medium (for example, a flexible disk) 34.

That is, the video signal photographed by the infrared camera 31 has an ultra-multi gradation (for example, 4096 gradations) and is taken into the microcomputer 32 through the input interface 321. The microcomputer 32 includes an input interface 321, a bus 322, and a CPU 3.
23, output interface 324 and memory 325
And executes the gradation conversion program stored in the external program recording medium 34.

That is, the gradation conversion program is recorded in the external program recording medium 34, and prior to gradation conversion, the gradation conversion program is loaded into the memory 325 via the input interface 321, and executed by the CPU 323. Then, the gradation-converted image is displayed on the display 33 via the output interface 324. The image conversion method according to the present invention may be incorporated in the infrared camera body to have the configuration shown in FIG.
In this case, the high gradation monitor 314 can be omitted.

That is, the infrared image passing through the lens system 311 is converted into an electric signal by the light receiving element 312 and amplified by the video amplifier 313. Amplified by video amplifier 313
The video signal is displayed on a high gradation monitor, and
It is input to the D converter 321. A / D converter 3
Reference numeral 21 has a 12-bit digital output capable of reproducing a super-multi-tone digital image.

The super multi-tone digital image output from the A / D converter 321 is sequentially taken into the memory 325 and subjected to tone conversion as described above. The gradation-converted image is displayed on the low gradation monitor 33, which is a normal display, via the output interface 324. When the microcomputer 32 is incorporated in the infrared camera body, the infrared camera becomes large and expensive if the microcomputer 32 has general versatility. Therefore, it is preferable that the microcomputer 32 is not a stored program system. Therefore, when the image conversion method according to the present invention is incorporated in the infrared camera body, the program is stored in the memory 325.
It is recorded from the beginning.

The CPU 323 may be composed of a digital signal processor or the like, and accordingly, the data bus 3
22 must be configured to correspond to a digital signal processor. Further, not only a method using a general-purpose processor and a program but also a circuit for realizing the present invention may be configured by combining a filter circuit and an edge extraction circuit configured by hardware.

FIG. 4 is a flow chart for explaining the principle of the gradation conversion processing according to the present invention. First, two adjacent pixels from the original image (A) 40 having M gradations have a predetermined threshold value. The edge emphasis processing for emphasizing the area having the above gradation difference is performed to obtain the edge emphasis image 41. Next, a pixel expansion process is performed on the edge-enhanced image 41 to create a mask pattern 42, and a mask 421 that allows only pixels around the edge (for example, three pixels centering on the edge) to pass through and pixels around the edge (for example, the edge An inversion mask 422 that allows pixels other than the central three pixels) to pass through is created.

Next, the original image (A) 40 is subjected to an edge-preserving noise removing process 43 for removing noise from the pixels around the covered edge with the edge peripheral pixel passing mask 421 while retaining the edge information and inverting the original image 40. An edge non-preserving noise removal process 44 for removing noise from pixels other than the periphery of the covered edge is performed by the mask 422, and a low frequency component image (D) 45 representing a global luminance change of the original image 40 is obtained.

The mask 421 or the reversal mask 4
It is also possible to determine the edge strength for each pixel of the original image without using 22, and select the noise removal processing method to be used. After that, gradation conversion processing for reducing the gradation of the low frequency component image 45 is performed, and the low frequency component gradation converted image 4
Create 5 '. Next, the low-frequency component image 45 is subtracted from the original image 40 to create a high-frequency component image 46 representing a local brightness change of the image, and then the brightness change is emphasized to create a high-frequency component brightness enhancement image 46 '. .

Then, the low-frequency component gradation converted image 45 'and the high-frequency component luminance-enhanced image 46' are added to each other to form a composite image 4
7 is created, and then the entire composite image 47 is converted into N gradations to obtain an output image 48. That is, the present invention is characterized in that the filter characteristics for obtaining the low frequency component image of the original image are changed according to the edge strength.

FIG. 5 is a first low frequency component extraction routine, and in step 50, the memory area reserved in the memory 325 for storing the edge emphasized image is cleared.
In step 51, the size m of the noise removal filter is set to a predetermined value m _MAX (for example, 25), and horizontal scanning is started for the original image (A) starting from the upper left pixel and ending at the lower right pixel. .

In step 52, it is determined whether or not an edge is included in the filter having the pixel at the center. If it is determined that the filter is not an edge portion, in step 53, an edge non-conservative noise removal filter of size m (eg, Noise removal is performed using an average value filter or a Gaussian filter). If it is determined in step 52 that it is an edge portion, the size m of the filter is decremented in step 54.
Next, in step 55, the filter size m is the minimum value m _{MI N}
It is determined whether or not (for example, 3), and if it is not the minimum value, the process returns to step 52 and repeats the processing.

When it is determined in step 55 that the edge is included even if the edge is reduced to the minimum value, in step 56, noise removal is performed using an edge-preserving noise removal filter (for example, median filter) of size m _MIN. carry out.
In step 57, it is determined whether or not the processing has been completed for all the pixels of the original image. If it is determined that the processing has not been completed, the process returns to step 51 to repeat the processing. When processing is completed for all pixels, low-frequency component extracted image (D)
Is completed and this process is completed.

FIG. 6 is a flowchart of a high frequency component extraction routine for creating a high frequency component image (E) by extracting a high frequency component representing a local luminance change from the original image (A). After clearing the memory area reserved in the memory 325 for storing (E), in step 61, the low frequency component image (D) is subtracted from the original image (A) for each pixel to obtain a high frequency component and a high frequency component is obtained. Store in component image (E).

In step 62, it is determined whether the high frequency components have been obtained over all the pixels of the original image (A), the processing is repeated until the high frequency components are calculated for all the pixels, and this routine is ended. FIG. 7 is a flowchart of a low frequency component gradation conversion routine for compressing the gradation of the low frequency component image (D) from M gradation to N gradation (M> N).
At 0, the memory area reserved in the memory 325 for storing the low frequency component gradation converted image (D ′) is cleared.

In step 71, for example, the gradation is converted for each pixel of the low frequency component image (D) by the following equation and stored in the gradation conversion image (D '). D ′ (i, j) = [N · {D (i, j) −M _MIN }] /
{M _MAX −M _MIN } where M _MAX is the maximum gradation of the original image (A) and M _MAX ≦ M
In addition, M _MIN is the minimum gradation of the original image (A) and M _MIN
≦ M _MAX .

In step 72, it is determined whether the gradation conversion has been performed on all the pixels of the low frequency component image (D), the processing is repeated until the conversion is completed for all the pixels, and then this routine is ended. FIG. 8 is a flowchart of a high frequency component enhancement processing routine for performing enhancement processing on all pixels of the high frequency component image (E). Cleared memory area.

In step 81, each pixel of the high-frequency component image (E) is subjected to an emphasis process based on the following equation, for example, and is stored in the emphasis process image (E '). E ′ (i, j) = n · E (i, j) (for example, n = 2) In step 82, it is determined whether emphasis processing has been performed on all the pixels of the high frequency component image (E), and conversion is completed for all the pixels. This routine is ended after repeating the processing until

FIG. 9 is a flow chart of a synthetic re-gradation conversion routine for obtaining the output image F ′.
At 0, the memory area reserved in the memory 325 for storing the composite image (F) and the output image (F ′) is cleared. In step 91, the gradation-changed image (D ′) and the emphasized image (E ′) are added for each pixel to create a composite image (F), and in step 92, the maximum brightness L _MAX and the minimum of the composite image (F) are obtained. Search for luminance L _MIN . Then, in step 93, the gradation of the composite image is reconverted for each pixel by the following equation to generate the output image (F ′).

F '(i, j) = [N {F (i, j)-
L _MIN }] / {L _MAX −L _MIN } In step 94, it is determined whether the re-gradation conversion processing has been performed on all the pixels of the composite image (F), and the processing is repeated until the conversion is completed for all the pixels. Exit the routine. Then, by outputting the output image (F ′) to the display 33 via the output interface 324, an image in which gradation around edges is not lost is displayed.

According to the above first gradation conversion method, it is determined for each pixel of the original image whether or not an edge is included in the noise removal filter centered on the pixel and used for noise removal. Since it is necessary to determine the filter, it is unavoidable that the processing amount becomes large. The second gradation conversion method is for solving this point, and the determination for each pixel is omitted by creating a mask in advance.

FIG. 10 is a flow chart of an edge emphasis processing routine for creating an edge emphasized image (B) which is a source of a mask from the original image (A), and in order to store the edge emphasized image (B) in step 100. The memory area secured in the memory 325 is cleared. Start horizontal scanning starting from the upper left for the pixels of the original image (A) and ending at the lower right,
In step 101, the edge strength of each pixel is calculated.

A known method can be used to calculate the edge strength. For example, when the Laplacian method is applied, the edge strength ES of the pixel A (i, j) of the original image (A) is calculated.
(I, j) is defined by the following equation. ES (i, j) = {f (i, j + 1) -f (i, j)}-{f (i, j) -f (i, j-1)} + {f (i + 1, j) -f (I, j)}-{f (i, j) -f (i-1, j)} = f (i, j + 1) + f (i, j-1) + f (i + 1, j) + f (i-1) , J) −4 · f (i, j) where f (i, j) represents the luminance of the pixel A (i, j).

In step 102, it is determined whether the edge strength ES is equal to or larger than a predetermined threshold value ε. If the edge strength ES is equal to or larger than the predetermined threshold value ε, in step 103, the pixel B ( Set the pixel value of i, j) to "1". On the contrary, if the edge strength ES is less than the threshold value ε, the pixel value of the pixel B (i, j) is set to "0" in step 104.

Then, in step 105, it is determined whether the edge enhancement processing has been completed for all the pixels of the original image (A), and if the processing of all pixels has not been completed, step 10
Return to 1. On the other hand, when the processing of all pixels is completed, this routine is ended. FIG. 11 is a flowchart of a pixel extension processing routine for creating a mask image (C) from the edge-enhanced image (B). In step 110, the number of times L of extension processing is repeated is set to a predetermined positive integer, and step 11
At 1, the memory area reserved in the memory 325 for storing the mask image (C) and the inverted mask image (C ′) is cleared.

After the edge-enhanced image (B) is transferred to the mask image (C) in step 112, the mask image (C) is horizontally scanned starting from the upper left and ending at the lower right.
At 3, it is determined whether the pixel value C (i, j) is "1". If the pixel value C (i, j) is "1", step 11
In 4 the pixel value is "8" near the pixel C (i, j).
It is determined whether there is a pixel that is "0", and the pixel value is "0".
If there is a pixel that is, the pixel expansion process for replacing the pixel value with "1" is executed in step 115, and the process proceeds to step 116. In step 113, the pixel value C (i, j) is "0".
, And in step 114 the pixel values in 8 neighborhoods are
If there is no pixel that is "0", the process directly proceeds to step 116.

FIG. 12 is an explanatory diagram of the pixel expansion processing.
"1" indicates that the pixel value is "1", and the diagonal line represents the pixel replaced by "1" as a result of the pixel expansion processing. Taking the pixel (7, 6) as an example, the 8 neighboring pixels are (6,
5), (7,5), (8,5), (6,6), (8,
6), (6,7), (7,7), (8,7),
Pixels (8,5) and (6,7) are originally pixel values "
Since it is 1 ", the pixel values of the other 6 pixels are changed from" 0 "to" 1 ".
By changing to, the pixel (7, 6) is expanded by one pixel.

In step 116, it is determined whether or not the processing has been completed for all pixels, and if the processing has not been completed for all pixels, the process returns to step 113 to repeat the processing. On the other hand, if the processing has been completed for all pixels, it is determined in step 117 whether the number of times L the extension processing is repeated is "1",
If L is not "1", the number of repetitions L is repeated in step 118.
Is decremented and the process returns to step 113.

If the number of repetitions L is "1", a reversed mask image (C ') in which the pixel value of the mask image (C) is reversed is created in step 119, and this routine is finished.
That is, when the number of repetitions L is set to "1", the edge is expanded by 1 pixel on both sides centering on the edge, and "2"
When set to, the pixel is expanded by 2 pixels. FIG. 13 is a flowchart of a second low frequency component extraction routine for creating a low frequency component image (D) by extracting a low frequency component representing a global luminance change from the original image (A).
Memory 325 for storing the low frequency component image (D) at
Clears the memory area secured within.

In step 131, the original image (A) is covered with the mask image (C), that is, {A × C} is calculated to extract the pixels around the edge from the original image (A), and in step 132, the pixels around the edge are extracted. Pixel, that is, the pixel value is "0"
Edge preservation noise removal processing for removing noise while retaining edge information is performed on non-pixels, and the result is stored in the low frequency component image (D).

A well-known filter can be used for the edge-preserving noise removal processing, but it is suitable to use a median filter. The median filter uses the median value (median) of the brightness of the pixels included in the area defined by the mask image (C), for example, 9 pixels included in the 3 × 3 area when the number of repetitions L = 1 as the brightness of the central pixel. It is a filter to do. Therefore, if the size of the median filter (when the number of repetitions L is 1, the size of the filter becomes “3”), not only the edge information becomes unclear but also the calculation time becomes long. In order to reduce the size, it is advantageous to set the number of repetitions L of mask making to be small.

In step 133, it is determined whether or not the median filter process for all the pixels of the mask image (C) is completed. If not completed, the process returns to step 132 to repeat the process, and if completed, the process proceeds to step 134. Step 1
In 34, the original image (A) is covered with the mask image (C ′),
That is, by calculating {A × C ′}, the original image (A)
Pixels other than the pixels around the edge are extracted from step S1, and edge non-conserved noise removal processing is performed in step 135 in order to reliably remove noise regardless of the edge information with respect to pixels other than the pixels around the edge that have non-zero brightness. Then, the result is stored in the low frequency component image (D).

A well-known filter can be used for the edge non-preserving noise removal processing, but it is suitable to use an average value filter or a Gaussian filter. It is advantageous to set the size of the filter as large as possible because it is not necessary to store the edge information in pixels other than the pixels around the edge and it is necessary to reliably remove noise.

In step 136, it is determined whether the average value filter or Gaussian filter process of all the pixels of the mask image (C) is completed. If not, step 1 is executed.
Returning to step 35, the process is repeated, and when it is finished, this routine is finished. That is, the second image conversion method generates a mask for extracting only pixels having a predetermined pixel width centering on an edge from the original image and an inversion mask for extracting only other pixels, and for the pixels extracted by the mask, By performing the edge-preserving noise removal processing and performing the edge-non-preserving noise removal processing on the pixels extracted by the inversion mask, the operation of determining the noise removal processing method for each pixel is omitted.

The process after the extraction of the low frequency components is the same as the first image conversion method. FIG. 14 is a functional diagram of the second gradation conversion method. In block 140, edge enhancement processing (FIG. 10) is performed from the original image (A) to generate an edge enhanced image (B), and block 141 In, the mask image (C) and the inverted mask image (C ′) are generated from the edge emphasized image (B) (FIG. 11).

In block 142, the original image (A) is covered with the mask image (C) to perform the edge-preserving noise removal processing.
In block 143, the original image (A) is covered with the mask image (C ′) to perform edge non-preserving noise removal processing (FIG. 8), and in block 144, the low frequency component image (D).
After obtaining, the tone conversion (FIG. 7) is performed in order to obtain the image (D ′) whose tone is compressed in block 145.

After the low frequency component image (D) is subtracted from the original image (A) in block 146 to calculate the high frequency component image (E) (FIG. 6), the high frequency component is emphasized in block 147 (FIG. 8). The obtained image (E ') is obtained. Then, in block 148, after obtaining a composite image (F) in which the low-frequency image (D ′) whose gradation is compressed and the high-frequency component image (E ′) that has been subjected to the emphasis process are combined, gradation compression is performed again. Tonal conversion is performed to obtain an output image (F ') (FIG. 9).

In the second gradation conversion method, the edge-preserving noise removal processing is performed on pixels that include edges, and the edge-non-preserving noise removal processing is performed on pixels that do not include edges. The process becomes complicated because it is necessary to install two types of filters for noise removal. The third gradation conversion method is for solving the above-mentioned drawback, and the size of the filter of the same format is set to be an edge peripheral pixel or a non-edge peripheral pixel without changing the filter format for noise removal. The program is simplified by changing it depending on

FIG. 15 is a flowchart of a third low frequency component extraction routine for generating a low frequency component image, which is reserved in the memory 325 to store the low frequency component image (D) in step 150. Cleared memory area. In step 151, the size m of the filter for noise removal processing (for example, average value filter, Gaussian filter) for noise removal is set to a predetermined positive integer value m ₀ (for example, 25), and then the original image (A) On the other hand, the horizontal scanning starts from the upper left and ends at the lower right.

Then, in step 152, it is judged whether or not there is an edge in the filter of size m ₀ , and if there is no edge, noise removal is performed by the filter of size m ₀ in step 153 and the result is low frequency. Store in component image (D). If it is determined in step 152 that there is an edge in the filter of size m ₀ , the size m of the filter is _decremented in step 154. In step 155, it is judged whether the filter size is reduced to "3",
If not reduced to "3", the process returns to step 152, and steps 152 and 154 are repeated until there are no edges in the filter or the size of the filter reaches "3".

In step 155, the filter size is changed to "
When the size reaches 3 ", the size is" 3 "in step 156.
The noise is removed by the filter (1) and stored in the low frequency component image (D). Then, in step 157, it is determined whether or not the processing has been completed for all areas. If not completed, the procedure returns to step 151, and if the processing for all areas is completed, this routine ends.

FIG. 16 is a functional diagram of the third gradation conversion method. In block 160, the low frequency component image (D) is extracted by the third low frequency component extraction routine. Grayscale conversion of the frequency component image (D) (block 16
1), extraction of high-frequency component image (E) (block 16
2), high-frequency component image (E) enhancement processing (block 16)
3), the combination process (block 164) and the re-tone compression conversion process (block 165) are the same as the first conversion method.

In the third gradation conversion method, noise can be removed by one type of filter, but the amount of processing increases because it is necessary to determine the size of the filter for each pixel. The fourth gradation conversion method uses a mask in the same manner as the second conversion method in order to solve the above-mentioned problems, so
The edge enhancement screen (B) is obtained by the same processing (FIG. 10) as the gradation conversion method of FIG.

FIG. 17 is a flowchart of the second pixel expansion processing routine for creating the second mask image (F) based on the edge-enhanced image (B).
In step 70, the number L of times of extension processing is set to a predetermined positive integer, and in step 171, the mask value P is set to "0". Then, in step 172, after clearing the memory area secured in the memory 325 for storing the second mask image (F), the edge emphasis screen (B) starts from the upper left and the horizontal scanning starts at the lower right. To do.

In step 173, it is determined whether or not the pixel value B (i, j) of the edge emphasis screen (B) is "1". If the pixel value B (i, j) is "1", step At 174, the pixel value F (i, of the corresponding pixel of the second mask image (F) is
Set j) to the mask value "P". However, "P" is 3
If it is less than "3". This is a measure for surely preventing an edge from entering the filter when performing noise removal processing on pixels near the edge.

In step 175, it is determined whether or not there is a pixel having a pixel value of "0" in the so-called 8 neighborhood of the pixel B (i, j). If there is a pixel having a pixel value of "0", step 1
At 76, the pixel value is replaced with "P". However, "P" is
If it is less than "3", it shall be "3". The reason for this is as described above. In step 173, the pixel value B
When (i, j) is "0", and step 175
If there is no pixel having a pixel value of "0" in 8 neighborhoods, the process directly proceeds to step 177.

In step 177, it is determined whether or not the processing has been completed for all pixels. If the processing has not been completed, the process returns to step 173 and the processing is repeated. On the contrary, if the process is completed, it is determined in step 178 whether the number of repeats L of the extension process is "1", and if L is not "1", the number of repeats L is decremented in step 179 to obtain the pixel value. Increment P and return to step 173. If L is "1", this routine ends.

FIG. 18 is a flow chart of a fourth low frequency component extraction routine used in the fourth tone conversion method. In step 180, the low frequency component image (D) is stored in the memory 325. After the secured memory area is cleared, the second mask image (F) is started from the upper left and the horizontal scanning is started from the lower right. In step 181, the size m of the noise removal filter is set to the pixel value, and in step 18
In step 2, the original image (A) is filtered using the filter, and the result is recorded in the low frequency component image (D).

In step 183, it is determined whether or not the processing has been completed for all pixels. If the processing has not been completed, the process returns to step 181 and the processing is repeated. On the contrary, if the processing is completed, this routine is ended. That is, in the fourth gradation conversion method, the pixel value of the pixel at a predetermined distance centered on the edge is set to "3" and the other pixel values are set to a predetermined value (for example, 2
The mask image (F) set in 5) is created, and the original image is subjected to noise removal by using a noise removal filter having the same size as the pixel value of the mask image (F), so that the filter size for each pixel is increased. The processing amount can be reduced by omitting the processing for determining

FIG. 19 is a functional diagram of the fourth conversion method. In block 190, the original image is changed to the edge enhancement screen (B).
Is obtained by the same method (FIG. 10) as the second conversion method, and then, in block 191, a second mask creating routine creates a second mask image (F) from the edge-enhanced image (B). Then, at block 192, the second mask image (F) is used to obtain the low frequency component screen (D) by the fourth low frequency component extraction routine.

Subsequent processing, gradation compression conversion of the low frequency component image (D) (block 193), high frequency component image (E)
Extraction (block 194), high-frequency component image (E) enhancement processing (block 195), and composition processing (block 19).
5) and the re-gradation compression conversion process (block 196) are the same as the second gradation conversion method. That is, according to the fourth gradation conversion method, it is not necessary to determine whether or not there is an edge in the filter for each pixel by using the mask image in which the size of the noise removal filter is set as the pixel value. Not only that, there is no need to change the type of noise removal filter, so the amount of calculation can be reduced significantly.

[0070]

According to the image gradation converting apparatus, the image gradation converting method and the medium for executing the method of the present invention, the portion (edge) having a large gradation difference in the super-multi gradation image is obtained. It is possible to reduce the gradation while retaining the information of No. 3, and it is possible to obtain a natural display without losing the gradation information near the edge even when the gradation information is displayed on a display of low gradation.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of a gradation conversion method.

FIG. 2 is an explanatory diagram of a problem.

FIG. 3 is a configuration diagram of an embodiment.

FIG. 4 is a flowchart of gradation conversion processing according to the present invention.

FIG. 5 is a flowchart of a first low frequency component extraction routine.

FIG. 6 is a flowchart of a high frequency component extraction routine.

FIG. 7 is a flowchart of a low frequency gradation conversion routine.

FIG. 8 is a flowchart of a high frequency component emphasis processing routine.

FIG. 9 is a flowchart of a synthetic gradation conversion routine.

FIG. 10 is a flowchart of an edge enhancement processing routine.

FIG. 11 is a flowchart of a pixel expansion processing routine.

FIG. 12 is an explanatory diagram of pixel expansion processing.

FIG. 13 is a flowchart of a second low frequency component extraction routine.

FIG. 14 is a functional diagram of a second gradation conversion method.

FIG. 15 is a flowchart of a third low frequency component extraction routine.

FIG. 16 is a functional diagram of a third gradation conversion method.

FIG. 17 is a flowchart of a second mask creation routine.

FIG. 18 is a flowchart of a fourth low frequency component extraction routine.

FIG. 19 is a functional diagram of a fourth gradation conversion method.

[Explanation of symbols]

31 ... Infrared camera 32 ... Microcomputer 321 ... Input interface 322 ... bus 323 ... CPU 324 ... Output interface 325 ... Memory 33 ... Display 34 ... Flexible disk

─────────────────────────────────────────────────── ─── Continued Front Page (58) Fields surveyed (Int.Cl. ⁷ , DB name) G06T 5/00-5/50 H04N 1/409 H04N 5/33

Claims (12)

(57) [Claims] 1. Separation means for separating an original image having super-multi-gradation into a low frequency component representing global gradation variation and a high frequency component representing local gradation variation, and between adjacent images of the original image. Edge strength detecting means for detecting the edge strength which is the gradation difference of the image, and changing the separation characteristic for separating the original image into the low frequency component and the high frequency component according to the edge strength detected by the edge strength detecting means. Separation characteristic changing means and reduction of gradation of low frequency components separated by the separating means
Gradation reducing means, the high frequency component separated by the separating means, and the gradation low
The low frequency component whose gradation is reduced by the reducing means is combined.
An image gradation conversion apparatus comprising: a synthesizing unit . 2. The separation characteristic changing means is a low frequency component extraction characteristic changing means for changing the low frequency component extraction characteristic according to the edge strength detected by the edge strength detecting means, and the separation means is A low-frequency component extraction means for extracting a low-frequency component from an original image by the low-frequency component extraction characteristic changed by the low-frequency component extraction characteristic changing means, and an original image and a low-frequency component extracted by the low-frequency component extraction means The image gradation conversion apparatus according to claim 1 , further comprising: a high frequency component extraction unit that extracts a high frequency component based on the above. 3. An edge peripheral pixel extraction means for extracting from the original image pixels within a predetermined number of pixels centered on a pixel having a high edge strength detected by the edge strength detection means. A non-edge peripheral pixel extracting means for extracting pixels other than the pixels extracted by the edge peripheral pixel extracting means from the original image;
Wherein the separation means performs edge preservation processing on the pixels extracted by the edge peripheral pixel extraction means and performs edge preservation processing on the pixels extracted by the non-edge peripheral pixel extraction means. The image gradation conversion apparatus according to claim 1, further comprising a low frequency component extraction unit that extracts a low frequency component. 4. The image floor according to claim 3 , wherein the low frequency component extraction unit reduces the size of the filter for extracting the low frequency component as the edge strength detected by the edge strength detection unit is higher. Key conversion device. 5. The edge surrounding pixel extraction means for extracting from the original image a pixel having a predetermined number of pixels centered on a pixel having a high edge strength detected by the edge strength detection means, and the separation characteristic changing means. A non-edge region extracting unit that extracts pixels other than the pixels extracted by the edge surrounding pixel extracting unit, wherein the separating unit filters the pixels extracted by the edge surrounding pixel extracting unit. The low frequency component is extracted by a filter having a small size, and the low frequency component is extracted by a filter having a large filter size for the pixels extracted by the non-edge peripheral pixel extraction means. Image gradation conversion device. 6. An infrared image pickup means capable of picking up an ultra-multi-gradation image, a low frequency component representing a global gradation variation of the image from the ultra-multi-gradation image, and a high frequency component representing a local gradation variation. Separating means for separating the, and the gradation of the low-frequency component separated by the separating means ,
Low-frequency component gradation conversion means for converting to a smaller number of gradations, and high-frequency component gradations separated by the separation means,
A high-frequency component emphasis processing means for emphasis processing, a low-frequency component gradation-converted by the low-frequency component gradation conversion means, and a synthesizing means for synthesizing the high-frequency component emphasized by the high-frequency component emphasis processing means. Characterized by
Infrared camera to be. 7. A separation step of separating an original image having super-multitone into a low frequency component representing global gradation variation and a high frequency component representing local gradation variation, and between adjacent images of the original image. The edge strength detection step of detecting the edge strength which is the gradation difference of, and the separation characteristic for separating the original image into a low frequency component and a high frequency component according to the edge strength detected in the edge strength detection step. Reduce the gradation of low frequency components separated by the separation characteristic change step and the separation step
Grayscale reduction step, the high frequency component separated by the separation step, and the grayscale low step.
Synthesize with low-frequency component whose gradation is reduced by the reduction step
An image gradation conversion method comprising: a combining step . 8. The separating characteristic changing step is a low frequency component extracting characteristic changing step of changing the low frequency component extracting characteristic according to the edge strength detected in the edge strength detecting step, and the separating step is A low frequency component extraction step of extracting a low frequency component from the original image by the low frequency component extraction characteristic changed in the low frequency component extraction characteristic changing step, and an original image and the low frequency component extracted by the low frequency component extraction step The image gradation conversion method according to claim 7 , further comprising a high-frequency component extraction step of extracting a high-frequency component based on. 9. The edge peripheral pixel extraction step of extracting from the original image a pixel within a predetermined number of pixels centered on a pixel having a high edge strength detected in the edge strength detection step in the separation characteristic changing step. And a non-edge region extracting step of extracting pixels other than the pixels extracted in the edge peripheral pixel extracting step from the original image, wherein the separating step is performed for the pixels extracted in the edge peripheral pixel extracting step. And a low frequency component extracting step of extracting a low frequency component from the pixel extracted in the non-edge peripheral pixel extracting step by the edge preserving processing.
7. The image gradation conversion method described in 7 . 10. The image floor according to claim 9 , wherein the low frequency component extracting step reduces the size of the filter for extracting the low frequency component as the edge strength detected in the edge strength detecting step increases. Key conversion method. 11. The edge peripheral pixel extraction step of extracting from the original image a pixel within a fixed number of pixels centered on a pixel having a high edge strength detected in the edge strength detection step in the separation characteristic changing step. And a non-edge peripheral pixel extraction step of extracting pixels other than the pixels extracted in the edge peripheral pixel extraction step , wherein the separation step is performed for the pixels extracted in the edge peripheral pixel extraction step. The image according to claim 7 , wherein the noise removal process is executed by a filter having a small filter size, and the pixel extracted in the non-edge peripheral pixel extraction step is executed by a filter having a large filter size. Tone conversion method. 12. The computer according to any one of claims 7 to 11.
A medium having recorded therein a program for executing the image gradation conversion method according to any one of items.