KR101024818B1 - Image processing apparatus using error diffusion method - Google Patents

Abstract

The present invention discloses an image processing apparatus using an error diffusion method. The image processing apparatus according to the present invention includes a detailed image map generator which receives an input image and generates a detailed image map representing a structural feature of the input image, and from the detailed image map generator, Receives a detailed image map, generates a threshold value necessary for binarizing a modified input image based on the values of the detailed image map, and binarizes and outputs pixel values of the changed input image according to the threshold value. A binarization unit, a subtractor for generating error values by subtracting the binarized value and the pixel value of the changed input image, and an error diffusion for generating a filtered error value by filtering the error values with its own weight An adder for generating the modified input image by adding the filtered error value to a pixel value of the input image; It includes.

Half-toning, error diffusion, detailed image map

Description

Image processing apparatus using error diffusion method

The present invention relates to an image processing apparatus using an error diffusion method according to shape information.

In a conventional image output apparatus such as a mono printer, a halftoning technique is mainly used to express gray scales of a multi-level that appears in an input image.

The half-toning technique outputs a low-bright part by increasing the number of dots, and outputs a high-bright part by reducing the number of dots, thereby producing a gray scale image. image).

Such half-toning techniques are classified into an ordered dither method and an error diffusion method.

First, the sequential dither method is a method of binarizing the grayscale value of each pixel using a fixed threshold value. In the sequential dither method, since the gray value of each pixel is determined uniformly according to a predetermined threshold value, the speed of determining the gray value is very fast, but there is a disadvantage in that image quality deteriorates due to the use of a fixed threshold value.

Next, the error diffusion method is a method of diffusing a quantization error due to binarization of pixels to surrounding pixels. In detail, the error diffusion method filters the binarization error generated after binarizing a pixel by using an error diffusion filter with a weight of a given filter, and adds the result to surrounding pixel values to update the original image to make the average error zero. The error diffusion method shows good performance in reproducing continuous grayscale images, but due to the inability to uniformly distribute binary pixels at a specific contrast value, it is distributed in the form of regular patterns or clusters, making it seem unobtrusive to human vision sensitive to low frequency components. . In addition, since the error diffusion filter is designed to maintain the average gray level of the original image to be zero, the binary image processed using this involves degradation of an edge region.

An object of the present invention for solving the above problems is to provide an image processing apparatus using an error diffusion method utilizing a fine structure and a small amount of image processing calculation.

In order to achieve the above object, the image processing apparatus according to an embodiment of the present invention receives an input image and generates a detailed image map representing structural features of the input image. A detailed image map generator, receiving the detailed image map from the detailed image map generator, generating a threshold required to binarize a modified input image based on values of the detailed image map, and A binarization unit for binarizing and outputting pixel values of the changed input image according to a threshold value, a subtractor for generating error values by subtracting the binarized value and pixel values of the changed input image, and generating the error values; An error diffusion filter for generating a filtered error value by filtering by its own weight, and the filtered error value in the pixel value of the input image By adding, an adder to generate the changed input image.

Here, the detailed image map generator generates a Laplacian map from the input image using a Laplacian filter.

Here, the image processing apparatus receives a detailed image map from the detailed image map generator, and has a structure weight to adjust the degree of structurality of the input image with respect to the values of the received detailed image map. and a weighting unit for assigning weight) (W [m, n]).

Here, the weighting unit applies an absolute sum to variance values in a predetermined window about each pixel of the detailed image map.

Here, the weighting unit sets the value of the detailed image map so that the range of the value of the detailed image map falls within the range of [0,255].

Here, the detailed image map generator generates a detailed image map by using the following equation.

Here, Inputaverage means the average of the input image I [m, n].

Here, the detailed image map generator generates a detailed image map by using the following equation.

Here, Gσ (input) is a value obtained by Gaussian filtering the input image, where σ is a Gaussian standard deviation. The sigma value can be appropriately adjusted according to the structure of the image.

Here, the detailed image map generator generates a detailed image map by using the following equation.

Here, B (input) is a value obtained by bilateral filtering the input image. In the case of σ _s , 20% of the smaller value of the width and height of the image is determined, and in case of σ _r (max intensity value minus the minimum intensity value) of the image / Is determined by 10.

According to another aspect of the present invention, there is provided an image processing apparatus, which receives an input image and generates a detailed image map representing a structural feature of the input image. Receiving a detailed image map from the map generator and the detailed image map generator, and generating a plurality of thresholds required for multi-toning a modified input image based on values of the detailed image map, A multi-toning unit which multi-tones and outputs pixel values of the changed input image according to the threshold values, and a subtractor which subtracts the multi-toned values and pixel values of the changed input image to generate error values; An error diffusion filter for filtering the error values by its own weight to generate a filtered error value, and a pixel value of the input image; And an adder for generating the modified input image by adding the filtered error value.

Here, the detailed image map generator generates a Laplacian map from the input image using a Laplacian filter.

Here, a structure-ness weight (W [m,) for receiving the detailed image map from the detailed image map generator and adjusting the degree of structurality of the input image with respect to the values of the received detailed image map. n]) further includes a weighting unit for assigning.

Here, the weighting unit applies an absolute sum to variance values within a predetermined window around each pixel of the detailed image map.

Here, the weighting unit sets the value of the detailed image map so that the range of the value of the detailed image map falls within the range of [0,255].

According to the image processing apparatus according to the present invention as described above, there is an effect of saving a small amount of calculation for the image processing and even a fine structure. In addition, according to the present invention, it shows a large effect in the RGB color image, not the binary image, and has a large effect in showing a small amount of computation for image processing and a fine structure even in multi-toning rather than half-toning. Applicable range is expected to be effectively used in display devices (PDP) and color printers, and because it is processed in real time, there is no problem to apply.

As the invention allows for various changes and numerous embodiments, particular embodiments will be illustrated in the drawings and described in detail in the written description.

However, this is not intended to limit the present invention to specific embodiments, it should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention. Like reference numerals are used for like elements in describing each drawing. It should be noted that the same elements in the drawings are represented by the same reference numerals and symbols as much as possible even though they are shown in different drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

Terms such as first, second, A, and B may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another.

For example, without departing from the scope of the present invention, the first component may be referred to as the second component, and similarly, the second component may also be referred to as the first component. And / or < / RTI > includes any combination of a plurality of related listed items or any of a plurality of related listed items.

When a component is referred to as being "connected" or "connected" to another component, it may be directly connected to or connected to that other component, but it may be understood that other components may be present in between. Should be. On the other hand, when a component is said to be "directly connected" or "directly connected" to another component, it should be understood that there is no other component in between.

The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting of the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, the terms "comprise" or "have" are intended to indicate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, and one or more other features. It is to be understood that the present invention does not exclude the possibility of the presence or the addition of numbers, steps, operations, components, components, or a combination thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art. Terms such as those defined in the commonly used dictionaries should be construed as having meanings consistent with the meanings in the context of the related art, and are not construed in ideal or excessively formal meanings unless expressly defined in this application. Do not.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. 1 is a block diagram of an image processing apparatus according to a first embodiment of the present invention.

Referring to FIG. 1, an image processing apparatus includes an adder 10, a binarizer 20, a subtractor 30, an error diffusion filter 40, and a Laplacian map generator 150.

The adder 10 receives an error value output from the input image and the error diffusion filter 40, adds the pixel value and the error value of the input image, and outputs a modified image to the binarizer 20. .

The binarizer 20 receives the changed image generated by adding the pixel value and the error value of the input image from the adder 10. The binarization unit 20 receives the Laplacian map from the Laplacian map generator 50.

The Laplacian map generator 50 receives an input image and generates a Laplacian map using a Laplacian filter from the input image. The size of the Laplacian filter is preferably 3 × 3. Laplacian filters are widely used for line detection, edge detection, and image enhancement techniques. The Laplacian filter is well known in the art and thus its detailed description is omitted.

The generated Laplacian map is used to extract structural features of the input image. That is, the values of the Laplacian map numerically represent the characteristics of the lines or edges in the input image. If the value of the Laplacian map is large, it is likely that the pixel value is a line or edge in the image. In addition, when the value of the Laplacian map is small, it is highly likely that the pixel value is a flat portion of the image, not a line or an edge.

The binarization unit 20 receives the Laplacian map and binarizes pixel values of the changed image received from the adder 10 based on the Laplacian map values.

In detail, the binarization unit 20 performs binarization by using different threshold values T [m, n] at flat portions and sharp portions, that is, edges or line portions. In other words, the binarization unit 20 determines threshold values used to binarize the changed image based on the values of the Laplacian map.

In an embodiment, when the value L [m, n] of the Laplacian map corresponding to the predetermined pixel value [m, n] in the image is greater than 127.5, the binarization unit 20 determines the threshold value T [m, n] as 255. When the value L [m, n] of the Laplacian map is 127.5 or less, the threshold value T [m, n] is determined to be 127.5. This is represented by Equation 1 below.

T [m, n] = 255, if L [m, n]> 127.5 (quick zone)

T [m, n] = 127.5, otherwise (flat area)

In another embodiment, when the value L [m, n] of the Laplacian map corresponding to the predetermined pixel value [m, n] in the image is greater than 32.5, the binarization unit 20 determines the threshold value T [m, n] as 255. When the value L [m, n] of the Laplacian map is 32.5 or less, the threshold T [m, n] is determined to be 127.5. This is represented by Equation 2 below.

T [m, n] = 255 if L [m, n]> 32.5 (quick zone)

T [m, n] = 32.5, otherwise (flat area)

In another embodiment, the binarization unit 20 determines the threshold value T [m, n] in proportion to the value L [m, n] of the Laplacian map. This is represented by equation (3).

T [m, n] = K (sharpness parameter) * L [m, n] + 127

Here, the sharpness parameter K may be adjusted to have a large value when trying to emphasize the structure of the image. And the constant K value is used to control the degree of structure of the image. In other words, the larger the K value, the higher the degree of structure-ness of the image. On the other hand, if the K value is less than or equal to 1, the degree of structure of the image is lowered.

As described above, the binarization unit 20 according to the present invention binarizes the pixel value of the changed image received from the adder 10 based on the values of the Laplacian map and outputs it to the subtractor 30.

The subtractor 30 subtracts the binarized value of the pixel value of the changed image and the pixel value of the changed image, thereby outputting error values to the error diffusion filter 40. The error diffusion filter 40 filters the calculated error value by its own weight and outputs the filtered error value to the adder 10. Accordingly, the adder 10 adds the filtering result of the error diffusion filter 40, that is, the filtered error value, to the pixel value of the original input image.

Accordingly, a binary image processed by half-toning using an existing error diffusion method involves deterioration of an edge region, and an image processed by an error diffusion processing apparatus according to the present invention exhibits better structure in an edge region. .

2 is a block diagram of an image processing apparatus according to a second exemplary embodiment of the present invention.

2, the image processing apparatus includes an adder 110, a binarizer 120, a subtractor 130, an error diffusion filter 140, a Laplacian map generator 150, and a weighting unit 160. ). The image processing apparatus illustrated in FIG. 2 has the same configuration except for the image processing apparatus and the weighting unit 160 according to the first embodiment. Therefore, the description of the components except for the weighting unit 160 refers to the description related to FIG. 1.

The adder 110 receives an error value output from the input image and the error diffusion filter 140, adds the pixel value and the error value of the input image, and outputs the modified image to the binarizer 120.

The binarizer 120 receives the changed image generated by adding the pixel value and the error value of the input image from the adder 110. The binarization unit 120 receives the Laplacian map from the Laplacian map generator 150.

The Laplacian map generator 150 receives an input image and generates a Laplacian map using a Laplacian filter from the input image. As described above, the generated Laplacian map is used to extract structural features of the input image. That is, the present invention allows the binarization to be performed using different threshold values in the flat area and the sharp area using only the Laplacian map L [m, n].

However, in order to solve this problem, the second embodiment of the present invention provides a structure-ness weight (W [] which is local spatial information. m, n]).

To this end, the error diffusion processing apparatus according to the second embodiment of the present invention includes a weighting unit 160 behind the Laplacian map generator 150.

The weighting unit 160 receives the Laplacian map from the Laplacian map generator 50, and compares the structure-ness weight (W [m, n) to the value L [m, n] of the received Laplacian map. ]). Accordingly, the threshold value used in the binarization unit 120 is calculated as in Equation 4 below.

T [m, n] = K (sharpness parameter) * W [m, n] * L [m, n] + T

Structure-ness weight W [m, n] may be applied according to the two methods described below.

The first method applies an absolute sum to variance values within a given window about each pixel in the Laplacian map L [m, n]. A large number of variance values of the Laplacian map in a given window means that the structure of the Laplacian filter results in a large number of structures.

In other words, the weighting unit 160 generates an absolute sum value by summing variance values within a predetermined window in order to reduce the calculation amount of variance. The use of 11 x 11 is to at least ignore the finest structures in the window.

3 shows the absolute sum of the Laplacian map. As shown in FIG. 3, the weighting unit 160 deteriorates the structure of the Laplacian map by obtaining an absolute sum for each value of the Laplacian map within a predetermined window area around the corresponding pixel.

The second method sets the value of the Laplacian map L [m, n] so that the range of the value of the Laplacian map L [m, n] falls within the range of [0,255]. That is, if the value of the Laplacian map (L [m, n]) exceeds 255, the weighting unit 160 sets the corresponding value to 255. Accordingly, the maximum value of the values of the Laplacian map (L [m, n]) is 255. The reason why the maximum value is 255 is to solve the problem that the structure-ness is emphasized too much when the value of the Laplacian map (L [m, n]) becomes too large and is greatly different from the surrounding pixels.

The binarization unit 120 receives the Laplacian map and binarizes pixel values of the changed image received from the adder 110 based on the Laplacian map values. In this case, the values of the Laplacian map are values to which the structural weight is applied in the weighting unit 160.

 The binarization unit 20 performs binarization using different threshold values T [m, n] at flat and sharp portions, that is, at edges or line portions.

4 is a block diagram of an image processing apparatus according to a third embodiment of the present invention.

Referring to FIG. 4, the image processing apparatus includes an adder 210, a multi-toning unit 220, a subtractor 230, an error diffusion filter 240, a Laplacian map generator 250, and a weighting unit ( 260). The image processing apparatus illustrated in FIG. 3 is almost similar except for the image processing apparatus, the multi-toning unit 220, and the Laplacian map generator 250 according to the second embodiment. The difference is that the input image is not a binary image but a color image. Therefore, the present embodiment will be described below with reference to the multi-toning unit 220.

First, the image processing apparatus according to FIG. 1 or 2 generates a threshold map T [m, n] to make an image binarized to 0 or 1 according to the half-toning method. The processed image processing apparatus generates a threshold map to produce a color image according to a multi-toning method, and the processed image well represents a fine structure.

5 is a diagram illustrating how a threshold is applied according to a halftoning method and a multitoning method. FIG. 5 (a) shows how the threshold is used according to the half-toning scheme and FIG. 5 (b) shows how the threshold is used according to the multitoning scheme.

Referring to FIG. 5B, the image processing apparatus according to the third embodiment of the present invention maps color values of pixels of an input image to 256 color values according to a multi-toning method.

According to the present embodiment, the multi-toning unit 250 maps color values of pixels of the input image to 256 color values and thus has 256 threshold values. The Laplacian map generator 250 generates a Laplacian map using a Laplacian filter from the input image and provides the Laplacian map to the multitoning unit 250. The multi-toning unit 250 determines a plurality of thresholds for multitoning the changed image based on the Laplacian map, and multitons the changed image according to the determined threshold values. Accordingly, the subtractor 230 subtracts the multi-toned value and the pixel value of the changed input image to generate error values.

6 is a block diagram of an image processing apparatus according to a fourth embodiment of the present invention.

The image processing apparatus illustrated in FIG. 6 has a difference in that the image processing apparatus illustrated in FIG. 2 and the detail image generator 320 are provided instead of the Laplacian map generator 150.

The Laplacian map is one of methods of extracting a detail image in which a structure exists, and accordingly, the portions 50 and 150 for obtaining a Laplacian map in FIG. 1 or 2 are detailed image images ( detail structure image).

Referring to FIG. 6, the image processing apparatus may include an adder 310, a binarizer 320, a subtractor 330, an error diffusion filter 340, a detailed image generator 350, and a weighting unit. unit) 360.

The adder 310 receives an error value output from the input image and the error diffusion filter 340, adds the pixel value and the error value of the input image, and outputs a modified image to the binarization unit 320.

The binarization unit 320 receives the changed image generated by adding the pixel value and the error value of the input image from the adder 310. The binarizer 120 receives the detailed image map from the detailed image generator 150.

The detailed image map generator 150 may be implemented according to various embodiments of the present disclosure.

According to an embodiment, the detailed image map generator 150 may employ a histogram equalization method.

In detail, when the detailed image map generator 150 receives the input image, the detailed image map generator 150 generates a histogram by graphing gray levels of the pixels included in the input image according to the frequency thereof. This histogram represents the characteristics of the image. For example, in the case of a dark image, most of the pixels included in the input image have a low gray level. In contrast, in the case of a bright image, most of the pixels included in the input image have a high gray level.

Subsequently, the detailed image map generator 150 generates a cumulative histogram from the histogram. That is, as the gray level increases, the number of pixels corresponding to each gray level accumulates. Next, the detailed image map generator 150 normalizes the cumulative histogram to perform histogram smoothing. Histogram equalization is an operation that improves the contrast, and is a method of improving the image quality of a low contrast image.

In this way, the detailed image map generator 150 generates a detailed image map having improved contrast by performing histogram smoothing on the input image.

According to another exemplary embodiment, the detailed image map generator 150 may generate the detailed image map using Equation 5 below.

Here, D [m, n] is a detailed image map, K is a sharpness parameter, and an input _average means an _average of the input image I [m, n].

Specifically, when the detailed image map generator 150 divides the input image by the average of the image, the gray level is displayed as 1 when the gray level is the same as the average image. . After all, it is determined that the portion other than 1 is a structure such as an edge or a line. At this time, subtracting 1 leaves only the structure part of the input image, and if it is about K times (about 1000), an image having only structure can be generated.

It is also possible to selectively apply a laplacian filter on the value of the resulting detail image map. If the detailed image does not represent the structure of the image, the image map indicating the structure may be generated by applying the Laplacian filter.

Alternatively, if the structure is too strong in the detailed image, that is, if there are too many structures in the detailed image, the structure-ness weight (Structure-ness weight) for reducing the structure with respect to the value L [m, n] of the detailed image map ( W [m, n]) can be given. As described above, in order to reduce the structure, the absolute sum is applied to the variance values within a predetermined window around each pixel of the detailed image map, or the value of the detailed image map L [m, n]. The value of the Laplacian map (L [m, n]) can be set so that the range of λ falls within the range of [0,255].

According to another embodiment, the detailed image map generator 150 may generate a detailed image map by using Equation 6 below.

Where G _σ (input) is a Gaussian filtering value of the input image, where σ is a Gaussian standard deviation. The sigma value can be appropriately adjusted according to the structure of the image.

According to this exemplary embodiment of the present invention, the detailed image map generator 150 performs Gaussain filtering to create a detail structure image. When Gaussian filtering is performed, the large scale structure of the input image remains and the small sacle structure is almost disappeared.

In this case, the difference from Equation 5 is divided into an average of the input image, that is, a constant value. In the case of Equation 5, when Gaussian filtering is performed according to Equation 6, an image is obtained. Divide by the local value of). After all, since the local gray level of the image (consideration) is considered, there is an advantage that the structure is better.

According to another embodiment, the detailed image map generator 150 may generate a detailed image map using Equation 7 below.

Here, B (input) is a value obtained by bilateral filtering the input image. The reason for the proposed bidirectional filtering method is that halo may occur in a large structure part of the detailed image through Gaussian filtering.

In the case of σ _s , 20% of the smaller value of the width and height of the image is determined.

In the case of σ _r , it is determined as (max intensity value min min value) / 10 of the image.

Although described above with reference to a preferred embodiment of the present invention, those skilled in the art will be variously modified and changed within the scope of the invention without departing from the spirit and scope of the invention described in the claims below I can understand that you can.

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

2 is a block diagram of an image processing apparatus according to a second exemplary embodiment of the present invention.

3 shows the absolute sum of the Laplacian map.

4 is a block diagram of an image processing apparatus according to a third embodiment of the present invention.

5 is a diagram illustrating how a threshold is applied according to a halftoning method and a multitoning method.

6 is a block diagram of an image processing apparatus according to a fourth embodiment of the present invention.

Claims (13)

In the image processing apparatus, A detailed image map generator for receiving an input image and generating a detailed image map representing a structural feature of the input image; Receiving the detailed image map from the detailed image map generator, generating a threshold required for binarizing a modified input image based on values of the detailed image map, and generating the changed input according to the threshold value A binarizer for binarizing and outputting pixel values of an image; A subtractor for generating error values by subtracting the binarized value and the pixel value of the changed input image; An error diffusion filter for filtering the error values with their own weights to produce a filtered error value; And an adder configured to generate the modified input image by adding the filtered error value to the pixel value of the input image. The method of claim 1, The detailed image map generator generates a Laplacian map from the input image using a Laplacian filter. The method of claim 1, A structure-ness weight (W [m, n]) for receiving a detailed image map from the detailed image map generator and adjusting a degree of structure of the input image with respect to values of the received detailed image map. The image processing apparatus further comprises a weighting unit for assigning. The method of claim 3, And the weighting unit applies an absolute sum to variance values in a predetermined window around each pixel of the detailed image map. The method of claim 3, And the weighting unit sets a value of the detailed image map so that a range of values of the detailed image map falls within a range of [0,255]. The method of claim 1, The detailed image map generation unit generates a detailed image map by using the following equation.

Here, D [m, n] is a detailed image map, K is a sharpness parameter, and an input _average means an _average of the input image I [m, n]. The method of claim 1, The detailed image map generation unit generates a detailed image map by using the following equation.

Where D [m, n] is a detailed image map, K is a sharpness parameter, G _σ (input) is Gaussian filtering the input image, and σ is a Gaussian standard deviation. The σ value can be adjusted according to the structure of the image. The method of claim 1, The detailed image map generation unit generates a detailed image map by using the following equation.

D [m, n] is a detailed image map, K is a sharpness parameter, and B (input) is a value obtained by bilateral filtering of the input image. In the case of σ _s , 20% of the smaller value of the width and height of the image is determined, and in case of σ _r (max intensity value minus the minimum intensity value) of the image / Is determined by 10. In the image processing apparatus, A detailed image map generator for receiving an input image and generating a detailed image map representing a structural feature of the input image; Receiving the detailed image map from the detailed image map generation unit, generates a plurality of thresholds required for multi-toning a modified input image based on the values of the detailed image map, and generates a plurality of thresholds A multi-toning unit for multitoning and outputting pixel values of the changed input image according to the present invention; A subtractor for generating error values by subtracting the multitoned value and the pixel value of the changed input image; An error diffusion filter for filtering the error values with their own weights to produce a filtered error value; And an adder configured to generate the modified input image by adding the filtered error value to the pixel value of the input image. 10. The method of claim 9, The detailed image map generator generates a Laplacian map from the input image using a Laplacian filter. 10. The method of claim 9, Structure-ness weight (W [m, n]) for receiving the detailed image map from the detailed image map generator and adjusting the degree of structurality of the input image with respect to the values of the received detailed image map. The image processing apparatus further comprises a weighting unit for giving. The method of claim 11, And the weighting unit applies an absolute sum to variance values in a predetermined window around each pixel of the detailed image map. The method of claim 11, And the weighting unit sets a value of the detailed image map so that the range of the value of the detailed image map falls within a range of [0,255].

Images