WO2006117919A1

WO2006117919A1 - Image processing method, image processing device and image processing program

Info

Publication number: WO2006117919A1
Application number: PCT/JP2006/303112
Authority: WO
Inventors: Sei Koh
Original assignee: Konica Minolta Holdings, Inc.
Priority date: 2005-04-28
Filing date: 2006-02-22
Publication date: 2006-11-09
Also published as: JPWO2006117919A1

Abstract

An image forming device is provided for performing high speed processing by using an edge preserving filter provided as hardware. The edge preserving filter used in the image forming device performs smoothing process by reducing image data to reduce the number of pixels and then returning the pixel number by enlarging process.

Description

Specification

Image processing method, image processing apparatus, and image processing program

The present invention relates to an image processing method, an image processing apparatus, and an image processing program for processing various types of image data, and in particular, at least the base component using an edge-maintaining filter for the detail component and the base component included in the image data The present invention relates to an image processing method, an image processing apparatus, and an image processing program. Background art

[0002] Image data includes a detail component in which the signal value of each pixel changes in small increments compared to the surrounding pixels, and a base component in which the signal value of each pixel does not change in small increments compared to the surrounding pixels. It is included. Here, the detail component represents a fine change in the image, while the base component represents a substantially uniform change in brightness within a certain range of the image.

Here, in the waveform diagram showing the signal value of the image data shown in FIG. 10, the solid line is the signal value of the image data, and the wavy line is the base component described above. That is, a detail component in which the signal value of each pixel changes in small increments compared to the peripheral pixels and a base component that shifts the signal value of the detail component by a certain amount are included.

[0004] It should be noted that in order to compress the dynamic range of image data for various types of image processing, after separating the base component and detail component of the image data and compressing only this base component, Again, many techniques are used to synthesize the base and detail components again.

[0005] By doing so, the amplitude of the signal value is large! /, While the base component is compressed but the detail component is not compressed. The detailed depiction and nuances of this are an advantage that will not be lost.

[0006] Note that, in order to separate the base component and the detail component as described above, the base component is regarded as a low-frequency component, and the detail component is regarded as a high-frequency component, and separation is performed based on the difference in the frequency component. Like that.

[0007] When performing the above-described separation, the base component is extracted from the target pixel. Many methods take average values of surrounding pixels by applying a filter of a certain size (smooth filter).

Here, as shown in FIG. 11 (a), image data in which a portion (edge portion) where the signal value of the base component changes rapidly is assumed. In such a case, when the base component is extracted using the smoothing filter described above, the edge portion included in the original base component (FIG. 11 (b) wavy line), as shown in FIG. 11 (b) solid line, In other words, the portion where the signal value of the base component changes suddenly is weakened.

[0009] Then, when the base component extracted with the edge portion blurred in this way is compressed (FIG. 11 (b)) and synthesized with the detail component, a “halo” as shown in FIG. 11 (c) is obtained. A component such as blurring or pale, or false contour, is generated.

[0010] It should be noted that the above halo is considered to be caused by the fact that the high-frequency component contained in the base component is not detected as the base component and thus is not compressed and affects the detail component side. ing.

[0011] And, in this way, as a solution method for preventing halo, the high frequency component can be recognized as the edge portion, and the base component can be extracted while maintaining the high frequency component of the edge portion. It is proposed to use a simple filter (hereinafter referred to as “edge preserving filter”).

FIG. 12 shows an operation example using the edge maintaining filter. Here, as shown in Fig. 12 (a), there is a part (edge part) where the signal value of the base component changes rapidly! Assume image data that speaks.

[0013] In such a case, when the base component is extracted using the edge preserving filter that is not used in the smoothing filter described in FIG. 11, the original base component (see FIG. 12B) is obtained. 12 (b) Similar to wavy lines), that is, the part where the signal value of the base component changes suddenly (edge part) is not blurred! /.

[0014] Then, when the base component (Fig. 12 (b)) extracted with the edge portion maintained in this way is compressed and combined with the detail component, as shown in Fig. 12 (c), It is possible to generate image data that is V-shaped due to coughing.

As a result, the detail component is compressed while compressing the base component having a large signal value amplitude. Since the image is not compressed, the brightness and other parts of the image are compressed, but the detailed description of the image is not lost.

[0016] Various methods have been proposed for such edge maintaining filters such as bilateral filters, ε filters, and median filters. The detailed contents of the bilateral filter as an example are described in Non-Patent Document 1 below.

[0017] Further, as an image process for obtaining the same result as the above-described edge maintenance filter, only a small amplitude component is smoothed and a smoothing process is not performed on a large edge component, thereby extracting a base. This technique is described in Patent Document 1 below.

Non-Special Reference 1: “Fast Bilateral Filtering for the Display of high-Dynamic- Range Imag esj Fredo Durand and Julie Dorsey, SIGGRAPH 2002

Patent Document 1: Patent Publication 2001-275015 (Page 1, Figure 1)

Disclosure of the invention

Problems to be solved by the invention

[0018] The calculation method of the bilateral filter (bilateral filter) in Non-Patent Document 1 described above can be expressed by the following equation (1).

[0019] [Equation 1]

^Js

^{^{fip ~ s) s (Ip ~}} Is -) (1 I) p

[0020] Here, k (s) is a normalization term, and f (p-s) is a smoothing filter that performs smoothing processing on a real space such as a Gaussian filter. Here, the Gaussian filter is a smoothing (blurring) filter having a normal (Gaussian) weighting function.

Note that the smoothing filter processing for performing the smoothing processing is schematically shown in FIG. That is, for each pixel, smoothing is performed with a weight of a normal distribution including peripheral pixels.

[0022] Further, g (Ip-Is) is a function called an influence function, which is a function of the difference between the peripheral pixel value Ip and the target pixel value Is. Here, the larger the difference is, the smaller the value of g is. It is prescribed as [0023] In general convolution, the image data and the filter data are converted to frequency space by Fourier transform and multiplied by power, and then returned to the real space by inverse Fourier transform to speed up processing. can do.

[0024] However, in the above equation (1), the value of g is made variable in the actual calculation, so the method using the Fourier transform cannot be directly applied.

[0025] Therefore, the pixel values are equally divided, a few are picked up, a set is made, each value is set as the target pixel value I, and the following equation (2) is calculated.

s

[0026] For example, pixel values 0 to 255 are equally divided as 0 to 31, 32 to 63, 64 to 95, 96 to 127, 128 to 159, 160 to 191, 192 to 223, 224 to 255, etc. To do. Here, an example in which the pixel value is 0 to 255 and the signal value width is 32 is divided into eight.

[0027] As a result, the value of g can be fixed, and smoothing processing on the frequency space by Fourier transform can be realized. Finally, the final output of pixel s is calculated by linear interpolation of the two output towns of ^ij closest to pixel I.

s

[0028] [Equation 2]

-(2)

[0029] However, in the Fourier transform, it is predicted that the circuit scale will increase and the memory capacity required in the process will increase. For this reason, there is a problem that it is difficult to implement the high-speed bilateral filter described in Non-Patent Document 1 described above as hardware in an actual device.

The present invention has been made in view of the above problems, and is an image processing method, an image processing apparatus, and an image processing capable of executing processing at high speed using a hardware edge maintenance filter. The purpose is to realize the program.

Means for solving the problem

That is, the present invention as means for solving the above-described problems is as described below.

[0032] (1) The invention described in claim 1 is an image including a detail component and a base component. An image processing method for extracting at least the base component from data, wherein the base component is extracted by recognizing a high frequency component of the image as an edge and performing smoothing processing while maintaining the edge. When extracting the base component using a possible edge-maintaining filter, the smoothing process is to reduce the number of pixels by reducing the image data and then returning the number of pixels by an enlargement process. This is an image processing method.

[0033] (2) The invention described in claim 2 is an image processing method for extracting at least the base component from image data including a detail component and a base component, wherein the high-frequency component of the image is extracted. When the base component is extracted using an edge maintaining filter that can extract the base component by recognizing the edge and executing smoothing processing while maintaining the edge, the smoothing processing is performed as follows: The image processing method is characterized in that the image data is decomposed into a plurality of frequency components, and the low frequency components among the decomposed components are retained and inversely transformed to reproduce the image data.

[0034] (3) In the invention according to claim 3, the edge preserving filter has an influence function g relating to a difference between the pixel of interest and its surrounding pixels, and a smoothing function for executing a smoothing process. 3. The image processing method according to claim 1, further comprising at least a filter f.

[0035] (4) The invention described in claim 4 is an image processing apparatus for extracting at least the base component from image data including a detail component and a base component, wherein the high-frequency component of the image is extracted. An edge preserving filter that extracts the base component by recognizing an edge and performing smoothing processing while maintaining the edge is provided, and the edge preserving filter reduces the number of pixels by reducing the image data. The image processing apparatus is characterized in that the smoothing process is executed by returning the number of pixels after an enlargement process.

[0036] (5) The invention described in claim 5 is an image processing apparatus for extracting at least the base component from image data including a detail component and a base component. An edge preserving filter that extracts the base component by recognizing an edge and performing smoothing processing while maintaining the edge is provided, and the edge preserving filter decomposes the image data into a plurality of frequency components, The smoothing process is performed by reserving and inversely transforming the low frequency component in the decomposed image and reproducing the image data. An image processing apparatus characterized by the above.

[0037] (6) In the invention according to claim 6, the edge maintaining filter includes an influence function part g relating to a difference between the pixel of interest and its surrounding pixels, and a smoothing process for executing a smoothing process. 6. The image processing apparatus according to claim 4, further comprising at least a filter part f.

[0038] (7) The invention according to claim 7 is an image processing program for extracting at least the base component from the image data including the detail component and the base component. An edge maintenance filtering routine for extracting the base component by causing a high frequency component to be recognized as an edge and performing a smoothing process while maintaining the edge is provided, and the smoothing process performs a reduction process on the image data. An image processing program characterized in that the number of pixels is reduced and the number of pixels is returned by enlargement processing.

[0039] (8) The invention described in claim 8 is an image processing program for extracting at least the base component from the image data including the detail component and the base component. An edge maintenance filtering routine for extracting the base component by recognizing a high frequency component as an edge and performing a smoothing process while maintaining the edge is provided, and the smoothing process converts the image data into a plurality of frequency components. This is an image processing program characterized in that it is a process of reconstructing image data and reconstructing the image data by suspending and inversely transforming the low frequency component in the decomposed data.

[0040] (9) In the invention according to claim 9, the edge maintenance filtering routine includes an influence function routine g relating to a difference between a pixel of interest and its surrounding pixels, and a smoothing that executes a smoothing process. The image processing program according to claim 7 or 8, further comprising at least a wrinkle filter routine f.

The invention's effect

In the present invention, the following effects can be obtained.

[0042] (1) In the invention of the image processing method according to claim 1 of the present invention, the edge maintaining filter is used to recognize the high frequency component of the image as an edge, and smoothing processing is performed while maintaining the edge. When the base component is extracted by execution, the image data is reduced as the smoothing process. Then, after reducing the number of pixels, return the number of pixels by enlargement processing!

[0043] In other words, since smoothing processing is reduced and expanded, it can be executed at high speed in real space, and since it does not depend on Fourier transform, large-scale processing of the circuit can be avoided and the process of processing can be avoided. This eliminates the need for memory to handle large temporary files, making it easier to create hard files. This makes it possible to execute processing at high speed using a hardware-based edge maintenance filter.

[0044] (2) In the invention of the image processing method according to claim 2 of the present invention, the high-frequency component of the image is recognized as an edge using the edge maintaining filter, and smoothing processing is performed while maintaining the edge. When the base component is extracted by execution, as this smoothing process, the image data is decomposed into a plurality of frequency components, and the low frequency components are retained in the decomposed and inversely converted to convert the image data. Reproduce.

[0045] That is, since smoothing processing is performed by inverse transformation of low frequency components obtained by frequency decomposition, it can be performed at high speed, and since it does not rely on Fourier transform, a large circuit is required. Scale can be avoided, and memory for handling large temporary files in the process is not required, making hardware easier. As a result, processing can be performed at high speed using a hardware-based edge maintenance filter.

[0046] (3) In the invention of the image processing method according to claim 3, in the above (1) or (2), the edge maintaining filter relates to a difference between the pixel of interest and its surrounding pixels. A bilateral filter having at least an influence function g and a smoothing filter f that executes smoothing processing can be used. Further, a trilateral filter having a function based on the gradient between the target pixel and its surrounding pixels can be used.

[0047] (4) In the invention of the image processing device according to claim 4, using the edge maintaining filter, the high frequency component of the image is recognized as an edge, and smoothing processing is performed while maintaining the edge. When the base component is extracted by execution, as the smoothing process, the image data is reduced and the number of pixels is reduced, and then the number of pixels is returned by the enlargement process.

[0048] That is, the smoothing process is reduced and expanded, so that it can be executed at high speed in real space, and since it does not rely on Fourier transform, a large-scale circuit is avoided and the process of processing is avoided. This eliminates the need for memory to handle large temporary files. It becomes easy to convert. This makes it possible to execute processing at high speed using a hardware-based edge maintenance filter.

[0049] (5) In the invention of the image processing device according to claim 5, using the edge maintaining filter, the high frequency component of the image is recognized as an edge, and smoothing processing is performed while maintaining the edge. When the base component is extracted by execution, as this smoothing process, the image data is decomposed into a plurality of frequency components, and the low frequency components are retained in the decomposed and inversely converted to convert the image data. Reproduce.

[0050] That is, since smoothing processing is executed by inverse transformation of low frequency components obtained by frequency decomposition, it can be executed at high speed, and since it does not rely on Fourier transformation, a large circuit is required. Scale can be avoided, and memory for handling large temporary files in the process is not required, making hardware easier. As a result, processing can be performed at high speed using a hardware-based edge maintenance filter.

[0051] (6) In the invention of the image processing device according to claim 6, in the above (4) or (5), the edge maintaining filter relates to a difference between the pixel of interest and its surrounding pixels. A bilateral filter having at least an influence function g and a smoothing filter f that executes smoothing processing can be used. Further, a trilateral filter having a function based on the gradient between the target pixel and its surrounding pixels can be used.

[0052] (7) In the invention of the image processing program according to claim 7, the high-frequency component of the image is recognized as an edge by using the edge maintaining filter, and the smoothness processing is performed while maintaining the edge. When the base component is extracted by executing the processing, as the smoothing process, the image data is reduced and the number of pixels is reduced and then the number of pixels is returned by the enlargement process.

[0053] In other words, since smoothing processing is reduced and expanded, it can be executed at high speed in real space, and since it does not depend on Fourier transform, large-scale processing of the circuit can be avoided and the process of processing can be avoided. This eliminates the need for memory to handle large temporary files, making it easier to create hard files. This makes it possible to execute processing at high speed using a hardware-based edge maintenance filter.

[0054] (8) In the invention of the image processing program according to claim 8, using the edge maintenance filter, the high frequency component of the image is recognized as an edge, and smoothing processing is performed while maintaining the edge. As the smoothing process is performed, the image data is decomposed into a plurality of frequency components, and the low frequency components in the decomposed are retained and inversely converted to perform image processing. To reproduce.

[0055] That is, since smoothing processing is executed by inverse transformation of low frequency components obtained by frequency decomposition, it can be executed at high speed, and since it does not rely on Fourier transformation, a large circuit is required. Scale can be avoided, and memory for handling large temporary files in the process is not required, making hardware easier. As a result, processing can be performed at high speed using a hardware-based edge maintenance filter.

[0056] (9) In the invention of the image processing program according to claim 9, in the above (7) or (8), the edge maintaining filter has an effect on the difference between the pixel of interest and its surrounding pixels. It becomes possible to use a bilateral filter having at least a function g and a smoothing filter f for performing smoothing processing. Furthermore, a trilateral filter having a function based on the gradient between the target pixel and its surrounding pixels can be used.

Brief Description of Drawings

FIG. 1 is a functional block diagram showing a configuration of an image processing apparatus according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing the configuration of a specific apparatus to which the image processing apparatus according to the embodiment of the present invention is applied.

FIG. 3 is a flowchart showing the operation of the image processing apparatus according to the first embodiment of the present invention (procedure of image processing method, procedure of image processing program).

FIG. 4 is an explanatory diagram showing a state of calculation at the time of processing image data in the operation of the image processing apparatus according to the first embodiment of the present invention (procedure of image processing method, procedure of image processing program).

FIG. 5 is a timing chart (characteristic diagram) showing how image data is processed in the operation of the image processing apparatus according to the first embodiment of the present invention (procedure of image processing method, procedure of image processing program). .

FIG. 6 is a timing chart (characteristic diagram) showing the state of processing of the main part in the operation of the image processing apparatus according to the first embodiment of the present invention (procedure of image processing method, procedure of image processing program).

). FIG. 7 is a timing chart (characteristic diagram) showing the state of processing of the main part in the operation of the image processing apparatus according to the first embodiment of the present invention (procedure of image processing method, procedure of image processing program).

).

[8] FIG. 8 is a functional block diagram showing the configuration of the image processing apparatus according to the second embodiment of the present invention.

FIG. 9 is an explanatory diagram showing the state of image data processing in the operation of the image processing apparatus according to the second embodiment of the present invention (procedure of image processing method, procedure of image processing program).

FIG. 10 is a characteristic diagram schematically showing the state of detail components and base components included in image data.

FIG. 11 is a timing chart (characteristic diagram) showing how base components are extracted using a smooth filter in conventional image processing.

FIG. 12 is a timing chart (characteristic diagram) showing how base components are extracted using an edge preserving filter in conventional image processing.

FIG. 13 is an explanatory diagram schematically showing a state of a Gaussian filter executed by an edge maintaining filter in conventional image processing.

Explanation of symbols

100 Image processing device

101 CPU

102 ROM

103 RAM

110 Calculation unit

120 1st filter part (edge preserving filter)

122 Frequency detector

124 Influence function calculator

140 Detail component extractor

150 Compression section

160 Synthesizer

170 Operation unit

180 Display 190 HDD

BEST MODE FOR CARRYING OUT THE INVENTION

Hereinafter, the best mode for carrying out the present invention (hereinafter referred to as an embodiment) will be described in detail with reference to the drawings.

First, a specific example of the electrical configuration of the first embodiment when the image processing method, the image processing apparatus, and the image processing program of the present invention are applied to the image processing apparatus will be described with reference to FIG.

That is, in the present embodiment, the operation of the image processing apparatus is the processing procedure of the image processing method, and the execution procedure of the image processing program is the operation of the image processing apparatus or the processing procedure of the image processing method. It is.

Here, the image processing apparatus includes the following means! FIG. 1 is a functional block diagram showing a functional configuration of a main part of the image processing apparatus according to the present embodiment, and FIG. 2 is a block diagram showing an example of specific circuits and apparatuses constituting the image processing apparatus. . The configuration of the image processing apparatus is an implementation means for executing the image processing method, and is also each routine of the image processing program.

Here, reference numeral 110 denotes an operation unit that executes image processing operations in the image processing apparatus 100 (see FIG. 2). This calculation unit 110 is in the image processing apparatus 100, and after separating the detail component included in the image data and the base component that shifts the signal value of the detail component by a certain amount, The calculation is performed again to generate image data.

[0064] It should be noted that this calculation unit 110 includes the edge portion while maintaining the edge portion correctly regardless of whether the edge portion in the base component included in the image data is abrupt or smooth. The base component can be extracted in the absence of the base component.

The arithmetic unit 110 functionally includes at least an edge maintaining filter 120, a detail component extraction unit 140, a compression unit 150, and a synthesis unit 160.

Here, reference numeral 120 denotes an edge maintaining filter that can extract a base component by executing smooth smoothing processing while maintaining an edge portion of an image and not including the edge portion. This edge The preserving filter 120 reduces the number of pixels by reducing the image data, and then returns the number of pixels by enlargement processing, thereby performing smoothing processing as a reduction / enlargement processing unit that executes smoothing processing in real space. Unit 122 and an influence function calculation unit 124 that calculates an influence function that outputs a value corresponding to a difference in signal value between the target pixel and the surrounding pixels (whether it is an edge or not).

Note that the edge maintaining filter 120 is a bilateral filter having at least an influence function _g relating to a difference between the target pixel and its surrounding pixels and a smoothing filter f that executes smoothing processing. Can be used. Further, it is possible to use an influence function _g , a smoothing filter f, and a trilateral filter having a function based on the gradient of the focused pixel and its surrounding pixels.

[0068] 140 subtracts the base component included in the input image data from the image data or divides the image data by the base component included in the input image data. It is a detail component extraction unit that extracts detail components included in the data. Here, the detail component is a component in which the signal value of each pixel of the image data changes in small increments compared to the surrounding pixels, and is mainly composed of a high frequency component.

[0069] Reference numeral 150 denotes a compression unit that performs a compression process for reducing the amplitude (dynamic range) by compressing the signal value of the base component extracted by the edge maintaining filter 120 at a predetermined ratio.

Reference numeral 160 denotes a synthesis unit that synthesizes the compressed base component and detail component. When the detail component extraction unit 140 performs subtraction, the synthesis unit 160 performs synthesis by adding the base component and the detail component. Further, when division is performed by the detail component extraction unit 140, the synthesis unit 160 performs synthesis by multiplying the base component and the detail component. The synthesizing unit 160 compresses the base component with a large amplitude of the signal value, but does not compress the detail component. Therefore, although the brightness and other parts of the image are compressed, the detailed depiction of the image is lost. The image data of the state is output.

In FIG. 2, the image processing apparatus 100 is configured to include the following as hardware: Has been.

[0072] The arithmetic unit 110 whose functional configuration is shown in Fig. 1 includes a CPU 101 as a control means for controlling each unit, a ROM 102 in which various data and various programs are stored, and image data is expanded and processed. RAM 103 as a work area.

[0073] 170 is an operation unit in which an operator (user) inputs various instructions to the image processing apparatus 100, 180 is a display unit that displays various states of the image processing apparatus 100, and 190 is image data and various data. HDD to be stored.

The operation of the image processing apparatus configured as described above, the processing procedure of the image processing method, and the execution procedure of the image processing program will be described below.

[0075] In this embodiment, the edge preserving filter 120 is used to perform the calculation of the above equation (2), and the influence function calculation unit 124 uses the influence function g to determine the difference between the pixel values and the high frequency of the image. The component is recognized as an edge, and the smoothing processing unit 122 executes a smoothing process while extracting the base component while maintaining the edge.

That is, the influence function calculation unit 124 monitors the difference in pixel value between the target pixel and the surrounding pixels using the influence function g. If the difference is large, it is an edge portion, and if the difference is small, it is a non-edge portion. Can be determined. In accordance with the calculation result of the influence function g of the influence function calculation unit 124, the smoothing processing unit 122 does not execute the smoothing processing at the edge portion but performs the smoothing processing at the non-edge portion. . In this way, the smoothing process is executed on the non-edge portion while maintaining the edge of the image. In such a smoothing process, in the present embodiment, as the smoothing process, after reducing the number of pixels by reducing the image data, the number of pixels is returned by an enlargement process.

[0077] That is, instead of using the conventionally proposed calculation in the frequency space by the Fourier transform, the smoothing processing by the smoothing processing unit 122 is temporarily reduced as shown in FIG. Decreasing the number of pixels and then enlarging it to increase (return) the number of pixels blurs the image, and a processing effect substantially equivalent to the smoothing filter processing can be obtained.

As an image reduction method in the smoothing processing unit 122, there is a method in which an image is divided into blocks, and an average value of each block is set as one pixel value of the reduced image. Also, for the enlargement method in the smoothing processing unit 122, the nearest four pixel forces are linearly interpolated. Law.

As described above, the smoothing processing unit 122 in the edge preserving filter 120 can be executed at high speed in the real space by changing the smoothing processing to reduction processing or enlargement processing. In addition, since the smoothing process does not rely on Fourier transform, a large-scale circuit can be avoided, a memory for handling a large temporary file in the process is unnecessary, and hardware is easy. become. This makes it possible to execute processing at high speed using a hardware-based edge maintenance filter.

[0080] FIG. 4 shows a calculation flow when the above-described processing of the first embodiment is applied to the above-described equation (2).

Here, the pixel value is equally divided, a few are picked up, a set Μ is created, each value is set as the target pixel value I, and equation (2) is calculated.

[0082] For example, the pixel values 0 to 255 are equally divided as 0 to 31, 32 to 63, 64 to 95, 96 to 127, 128 to 159, 160 to 191, 192 to 223, 224 to 255, etc. To do. Here, in the case of pixel values of 0 to 255, the signal value width 32 may be divided into eight other divisions.

[0083] That is, in the divided section, J is in the range of 0 to NB_SEGMENTS, G is calculated by the influence function g, the normalization coefficient K is calculated, H is calculated from G and the pixel value I, This H is smoothed by reduction and enlargement. Further, J is obtained by normalizing the smoothed value with a normalization coefficient K. Finally, the final output Js of pixel s is calculated by linear interpolation of the two output cities ^j of ^ij closest to pixel Is.

In the above description of the first embodiment, the effect of the smoothing process by the reduction process and the enlargement process is affected by the reduction / enlargement magnification.

[0085] If the image size force after reduction by reducing the enlargement magnification larger than appropriate is reduced too much, the smoothness value of the image becomes the value of the neighboring pixels close to each other as shown in FIG. 5 (b). It is greatly affected by the high frequency components, and as a result, the edge portion may be weakened. Then, a state called a mouth is generated in this part. Therefore, a good quality image cannot be obtained.

[0086] On the other hand, if the image size after reduction is too small by reducing the enlargement / magnification magnification at the appropriate time, the smoothing process is executed in a range that is too wide as shown in FIG. 6 (b). The dynamic range of the extracted base value becomes smaller. For this reason, the base component The change remains and the dynamic range of the detail increases, and the range of the final composited image also increases (see Figure 7). Therefore, the original purpose cannot be achieved.

Therefore, it is desirable to calculate the image size after reduction according to the following equation (3) based on experience values.

[0088] a 1 X image width Z image height <reduced size <«2 image width Z length…)

Here, α 1 = 0.01 and α 2 = 0.1. If the calculation result is not an integer, it is rounded to the nearest integer.

[0089] With such a size, that is, by performing a smoothing process by performing a reduction / enlargement process at a magnification, a more optimal edge maintenance smoothing process effect can be obtained with any image size.

[0090] As a modification, the reduction and enlargement method is not limited to that described in the first embodiment.

Any method may be used, such as a two-arrest neighbor method or a bicubic method. As a result, it becomes possible to deal with the diversity of processing.

[0091] <Second Embodiment>

A second embodiment in which the image processing method, the image processing apparatus, and the image processing program of the present invention are applied to the image processing apparatus will be described.

Here, in this embodiment, the operation of the image processing apparatus is a processing procedure of the image processing method, and the execution procedure of the image processing program is the operation procedure of the image processing apparatus or the processing procedure of the image processing method. is there.

[0093] In the second embodiment, the smoothing process in the smoothing processing unit 122 in the first embodiment described above is performed in place of the calculation in the frequency space by the conventional Fourier transform, and the first embodiment. As shown in FIG. 8, a smoothing processing unit 123 serving as a frequency resolving unit is arranged and a frequency resolving method is used instead of the form reducing / enlarging process.

Here, the smoothing processing unit 123 decomposes the image data into a plurality of frequency bands as shown in FIG. 9 (FIGS. 9 (a), (b), (c)) and generates the lowest frequency. Only the components are retained (Fig. 9 (d)) and reversely converted (Fig. 9 (e)). By doing so, the image is blurred, and a processing effect substantially equivalent to the smoothing filter processing can be obtained. For the frequency resolution method, wavelet transform can be used! As described above, the smoothing processing unit 123 in the edge preserving filter 120 can be executed at high speed by performing the frequency decomposition / combination processing for the smoothing processing. In addition, since the smoothing process does not rely on Fourier transform, a large-scale circuit can be avoided, and a memory for handling a large temporary file in the process is not necessary, making hardware easy. Become. As a result, it is possible to execute processing at high speed using a hardware edge preserving filter.

[0096] In the above description of the second embodiment, the effect of the smoothing process by the frequency decomposition process' synthesis process is affected by the decomposition factor of the frequency decomposition process.

That is, if the image resolution after frequency decomposition becomes too small by increasing the frequency resolution magnification from the appropriate value, the smoothness value of the image is a neighboring pixel close to the distance as shown in FIG. 5 (b). It is greatly affected by the high-frequency component of, and as a result, the edge portion may be blurred. And a state called a mouth is generated in this part. Therefore, a good quality image cannot be obtained.

[0098] On the other hand, if the frequency resolution magnification is made smaller than when it is appropriate and the image size after frequency resolution is too large, smoothing processing is performed in a range that is too wide as shown in Fig. 6 (b). The dynamic range of the extracted base value becomes smaller. As a result, changes in the base component remain in the detail, increasing the dynamic range of the detail and eventually increasing the range of the synthesized image (see Figure 7). Therefore, the original purpose cannot be achieved.

[0099] Therefore, it is desirable to calculate the frequency decomposition magnification that determines the image size after frequency decomposition based on empirical values as shown in the following equation (4). log (a 1 X image width Z length)

2

<Decomposition level log (α 2 ΧImage width ΖLength) · · · · (4),

2

Here, α 1 = 0.01 and α 2 = 0.1. If the calculation result is not an integer, it is rounded to the nearest nonzero integer.

[0100] By executing smoothing processing by frequency resolution processing and frequency synthesis processing at such a size or magnification, a more optimal edge maintenance smoothing processing effect can be obtained for any image size. It is done.

[0101] The image data is once decomposed into a plurality of frequency bands by the frequency decomposition method, and the lowest frequency In addition to the wave number component, other desired low frequency components may be reserved and inversely converted. In this way, it becomes possible to deal with the diversity of purposes.

In addition, each of the above embodiments can be applied to color image data including only monochrome image data. When applied to color image data, it can be applied to each color-specific image data such as RGB or Y MCK. Further, in the case of a luminance signal 'color difference signal or the like, it can be applied to only the luminance signal or to V and deviation of both the luminance signal and the color difference signal.

Claims

The scope of the claims

[1] An image processing method for extracting at least the base component from image data including a detail component and a base component,

By recognizing the high frequency component of the image as an edge and performing smoothing processing while maintaining the edge, the base component can be extracted using the edge maintaining filter that can extract the base component. An image processing method comprising: reducing the number of pixels by reducing the image data and then returning the number of pixels by an enlargement process as the smoothing process.

[2] An image processing method for extracting at least the base component from image data including a detail component and a base component,

By recognizing the high frequency component of the image as an edge and performing smoothing processing while maintaining the edge, the base component can be extracted using the edge maintaining filter that can extract the base component. As a smoothing process, the image data is decomposed into multiple frequency components.

An image processing method characterized in that the low frequency component in the decomposed state is retained and inversely transformed to reproduce the image data.

[3] The edge preserving filter has an influence function g on the difference between the target pixel and its surrounding pixels, and

The image processing method according to claim 1, further comprising: a smoothing filter f that executes a smoothing process.

[4] An image processing apparatus for extracting at least the base component from image data including a detail component and a base component,

An edge maintaining filter that extracts the base component by recognizing the high frequency component of the image as an edge and performing smoothing processing while maintaining the edge,

The image processing apparatus according to claim 1, wherein the edge preserving filter performs the smoothing process by reducing the number of pixels by reducing the image data and then returning the number of pixels by an enlargement process.

[5] An image processing apparatus for extracting at least the base component from image data including a detail component and a base component,

An edge maintaining filter that extracts the base component by recognizing the high frequency component of the image as an edge and performing smoothing processing while maintaining the edge, The edge preserving filter performs the smoothing process by decomposing the image data into a plurality of frequency components, reserving the low frequency component in the decomposed and performing inverse transformation to reproduce the image data. An image processing apparatus.

[6] The edge preserving filter has at least an influence function part g related to a difference between a pixel of interest and its surrounding pixels, and a smoothing filter part f that executes a smoothing process. The image processing apparatus according to item 4 or 5.

[7] An image processing program for extracting at least the base component for image data including a detail component and a base component,

An edge maintaining filtering routine for extracting a base component by causing a computer to recognize a high frequency component of an image as an edge and performing a smoothing process while maintaining the edge;

The smoothing process is a process for reducing the number of pixels by reducing the image data and then returning the number of pixels by an enlargement process.

[8] An image processing program for extracting at least the base component from image data including a detail component and a base component,

The smoothing process is an image processing program characterized in that the image data is decomposed into a plurality of frequency components, and the low frequency components among the decomposed components are retained and inversely transformed to reproduce the image data. .

[9] The edge maintaining filtering routine includes at least an influence function routine _g regarding a difference between a pixel of interest and its surrounding pixels and a smoothing filter routine f for executing smoothing processing. The image processing program according to claim 7 or 8.