JP2013004094A - Text emphasis method and device and text extraction method and device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method of making a computer emphasize text in an image.SOLUTION: Disclosed is the method of making a computer execute: a current image acquisition step of acquiring an original image including text of at least one line; an update value acquisition step of performing stroke two-dimensional filtering to an original luminance value or/and original color value of each original pixel point on the basis of the degree of direct difference and the degree of indirect difference from an arbitrary original pixel point in the original image to each neighboring pixel point in its neighboring aggregate, and acquiring an updated luminance value or/and updated color value after the filtering of the original image; and an emphasized image generation step of replacing the corresponding original luminance value or original color value with the updated luminance value or/and updated color value after the filtering of the original image, and generating a text emphasis image corresponding to the original image. The range of the neighboring aggregate is shaped like a square having the original pixel point as a center and a side length of w, in which w is smaller than the height of the original image.

Description

  The present invention relates to image processing, and more particularly, to a method and apparatus for enhancing text in an image, and a method and apparatus for extracting text in an image.

  During the process of playing a video or when viewing an image, the video or image often includes a text description. The text description is, for example, an explanation of the occurrence time and place of an event in the video, an interpretation of an image, or the like. Since these text contents are closely related to the video or image, the extraction of text in the video or image becomes a very important technique.

  In the prior art, methods for extracting video or image text can extract text in an image or video based on binarization or edge color clustering and detection techniques.

  However, when extracting text with the prior art, the video or image is too noisy, the image or video is unclear, and thus there is a lighting change in part of the video, so the text in the image or video The boundary between the background and the background becomes unclear, or the text content becomes unclear, which affects the text extraction effect.

  Therefore, how to perform enhancement processing on the text in the original image or video to enhance the text in the image or video and further optimize the effect of text extraction should be solved by the prior art. It becomes a problem.

  In view of the above-mentioned problems, the object of the present invention is for text in an original image containing at least one line of text so that the text in the original image becomes clearer and the effect of text extraction can be optimized. It is an object of the present invention to provide a text emphasizing method and apparatus that can perform emphasis processing, and a text extracting method and apparatus.

  According to one embodiment of the present invention, a method is provided for a computer to enhance text in an image. In this method, the computer obtains an original image including at least one line of text, a direct difference degree and an indirect difference from an arbitrary original pixel point in the original image to each neighboring pixel point in the neighborhood set. An update value for performing two-dimensional handwriting filtering on the original luminance value or / and primary color value of each original pixel point based on the degree and obtaining an updated luminance value or / and an updated color value after filtering of the original image And obtaining the text-enhanced image corresponding to the original image by replacing the corresponding original luminance value and / or primary color value with the updated luminance value or / and the updated color value after filtering of the original image, respectively. The enhanced image generation step is executed. The range of the neighborhood set is a square centered on the original pixel point and having a side length of w, and w is smaller than the height of the original image.

  According to the embodiments of the present invention, it is possible to provide a method and apparatus for enhancing text in an image and a method and apparatus for extracting text in an image.

4 is a flowchart illustrating a first type of text enhancement method according to an embodiment of the present invention. It is a flowchart which shows step S102 in the 1st type text emphasis method. 6 is a flowchart illustrating a second type of text enhancement method according to an embodiment of the present invention. It is a flowchart which shows step S302 in the 2nd type text emphasis method. It is another flowchart which shows step S302 in the 2nd type text emphasis method. It is another flowchart which shows step S302 in the 2nd type text emphasis method. It is another flowchart which shows step S302 in the 2nd type text emphasis method. It is a flowchart which shows step S304 in the 2nd type text emphasis method. 1 is a schematic diagram illustrating a first type of text emphasis device according to an embodiment of the present invention. It is a schematic diagram which shows the filter module 902 in a 1st type text emphasis apparatus. It is a schematic diagram which shows the 2nd type text emphasis apparatus by one Example of this invention. It is a schematic diagram which shows the stroke polarity estimation module 1101 in a 2nd type text enhancement apparatus. It is another schematic diagram which shows the stroke polarity estimation module 1101 in a 2nd type text enhancement apparatus. It is another schematic diagram which shows the stroke polarity estimation module 1101 in a 2nd type text enhancement apparatus. It is another schematic diagram which shows the stroke polarity estimation module 1101 in a 2nd type text enhancement apparatus. It is a schematic diagram which shows the judgment module 1102 in a 2nd type text emphasis apparatus. 3 is a flowchart illustrating a text extraction method according to an embodiment of the present invention. 1 is a schematic diagram illustrating a text extraction apparatus according to an embodiment of the present invention. It is a block diagram which shows the typical structure of the computer as an information processing apparatus by one Example of this invention.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a flowchart illustrating a first type of text enhancement method according to an embodiment of the present invention. As shown in FIG. 1, the first type of text enhancement method specifically includes the following steps. The first type of text enhancement method may be executed by a computer, for example.

  S101: An original image including at least one line of text is acquired.

  In the embodiment of the present invention, the above-described text emphasis refers to performing emphasis processing on text in an original image including at least one line of text. It can be understood that the emphasis here is to deepen the edges of the text or to distinguish the text from the background. Also, embodiments of the present invention, when applied, emphasize the consistency of pixels within a stroke and at the same time achieve the effect of deepening the difference between text and background (e.g. luminance or Color) and its shape (for example, text becomes a striped pattern) information.

  S102: Based on a direct difference degree and an indirect difference degree from an arbitrary original pixel point in the original image to each neighboring pixel point in the neighborhood set, the original luminance value or / and the primary color value of each original pixel point are obtained. On the other hand, stroke two-dimensional filtering is performed to obtain an updated updated luminance value and / or updated color value after filtering of the original image. The range of the neighborhood set is a square whose center is the original pixel point and whose side length is w, and w is smaller than the height of the original image.

  The direct difference degree in this step indicates a direct appearance difference, for example, a color or luminance difference, from an original pixel point to each region pixel point in the neighborhood set, and an indirect difference degree is determined from each original pixel point in the neighborhood set. The gradient module of the pixel that has passed through the region pixel point is shown. The updated brightness value or / and the updated color value after the original image is filtered by performing the two-dimensional filtering on the original brightness value and / or the primary color value of each original pixel point according to the direct difference level and the indirect difference level. Can be obtained. The neighborhood set here can be acquired so as to take a square centered on the original pixel point and having the side length of w. Note that w is preferably smaller than the height of the original image and is 1/8 of the height of the original image.

  FIG. 2 is a flowchart showing step S102 in the first type of text enhancement method. In actual application, as shown in FIG. 2, the above-described S102 specifically includes the following contents.

  S201: Algebraic subtraction is performed on the original luminance value or / and the primary color value between the original pixel point and each neighboring pixel point to directly obtain the degree of difference.

In this embodiment, when the direct difference degree between the pixels i and j is indicated by D ₁ (i, j), the direct difference degree of the luminance value can be calculated by the following equation (1).

  However, f ′ (i) represents the average value of the brightness of each neighboring pixel in the neighborhood set of the target pixel, that is, the difference value is directly calculated using the average value of the pixel values of the neighboring pixels instead of the pixel value of the target pixel. calculate. σ (i) indicates the local luminance standard deviation around pixel i and can be useful for normalization.

Further, the direct difference degree between the luminance values of the pixels i and j can be calculated by the following equation (2).

  Here, f (i) indicates the average value of the luminance of each neighboring pixel in the neighborhood set of target pixels.

Here, it should be explained that when calculating the direct difference between the color values of the pixels i and j, the following equations (3) or (4) can also be employed.

  N in the equations (3) and (4) indicates an R channel, a G channel, and a B channel as pixel color information. It should be explained here that the above-described formula for calculating the direct difference is merely illustrative. One skilled in the art can make adaptive changes to the above equations.

  S202: An indirect difference is acquired based on a gradient module from the original pixel point to each neighboring pixel point in the neighborhood set.

In this embodiment, when the indirect difference degree between the pixels i and j is indicated by D ₂ (i, j), the indirect difference degree between the luminance values of the pixels i and j can be calculated by the following equation (5). it can.

は、iからjへの方向に沿う、画素lにおけるグラジエントモジュールを示す。式（５）におけるbと式（１）におけるaはともに予め設定された所定のパラメータである。フィルタリングの平滑程度を制御するように両者の単調性が一致している。 However,

Indicates the gradient module at pixel l along the direction from i to j. Both b in the equation (5) and a in the equation (1) are predetermined parameters set in advance. Both monotonicities match to control the smoothness of filtering.

Of course, in an actual application, the difference between the maximum and minimum gradient values of the pixel luminance from i to j can be calculated instead of the maximum gradient value in Equation (5). The calculation method is shown in the following formula (6).

  Here, Max indicates the maximum value of the gradient value, and min indicates the minimum value of the gradient value.

In addition, the degree of indirect difference between the color values of the pixels i and j can be calculated by the following equations (7) and (8), respectively.

  n represents an R channel, a G channel, and a B channel as pixel color information. It should be explained here that the above-described formula for calculating the indirect difference is merely illustrative. One skilled in the art can make adaptive changes to the above equations.

  S203: Based on the direct difference degree and the indirect difference degree, a luminance value and / or a color value weight value of each neighboring pixel point with respect to the original pixel point is calculated.

After obtaining the indirect difference degree and the direct difference degree, the weight value can be calculated by the following equation (9).

Where D ₁ (i, j) indicates the direct difference between the luminance values and / or color values of the pixels i and j, and D ₂ (i, j) indicates the indirect difference between the luminance values and / or the color values of the pixels i and j. Degrees. Note that w (i, j) represents a luminance weight value, and w _n (i, j) represents a color value weight value.

S204: The updated luminance value of the original pixel point is calculated by the following stroke two-dimensional filter equation (10).

  Here, N (i) indicates a neighborhood set of the pixel point i, and w (i, j) indicates a weight value of the luminance value of the neighborhood pixel point j with respect to the original pixel point i. f (j) is the luminance value of the pixel point j in the neighborhood set.

S205: The updated color value of the original pixel point is calculated by the following stroke two-dimensional filter equation (11).

Where w _n (i, j) represents the weight value of the color value of the neighboring pixel point j relative to the original pixel point i in the n channel, and f _n (j) represents the color of the pixel point j in the neighborhood set in the n channel. Value.

  It should be explained here that steps S204 and S205 calculate and execute an updated luminance value and an updated color value, respectively, so that any one step is selected and executed in an actual application, or two steps are performed simultaneously. By executing, the present embodiment can be realized.

  Refer to FIG. 1 again.

  S103: Replace the corresponding original luminance value or / and primary color value with the filtered updated luminance value or / and updated color value, respectively, to generate a text enhanced image corresponding to the original image.

  After obtaining the updated luminance value or / and the updated color value after filtering, the primary luminance value or / and the primary color value are replaced with the updated luminance value or / and the updated color value, respectively. After such replacement, the text stroke at the pixel points in the original image is emphasized, and the matching of the pixels in the stroke is enhanced, and the degree of difference between the text and the background is deepened.

  FIG. 3 is a flowchart illustrating a second type of text enhancement method according to an embodiment of the present invention. As shown in FIG. 3, the second type of text enhancement method specifically includes the following steps. Note that this second type of text emphasis method may be executed by a computer, for example.

  S301: An original image including at least one line of text is acquired.

  S302: The stroke polarity of the text in the original image is estimated. The polarity indicates the relationship between the magnitude of the luminance value and / or the color value between the pixel point located inside the stroke area and the pixel point located outside the stroke area.

  In actual application, text stroke enhancement mainly uses filter technology, that is, a pixel value outside the stroke to emphasize the target pixel value in the stroke, so the area around the target pixel point in the stroke. The noise pixel point of this will adversely affect the effect of stroke enhancement. This effect becomes more apparent when processing thin strokes or stroke intervals. In order to avoid such a drop in quality, stroke polarity estimation is introduced in this embodiment. The stroke polarity estimated in this step can indicate the relationship between the magnitude of the luminance value and / or the color value between the pixel point inside the stroke area and the pixel point outside the stroke area.

  FIG. 4 is a flowchart showing step S302 in the second type of text enhancement method. Specifically, when the polarity indicates the relationship of the magnitude of the luminance value between the pixel point inside the stroke area and the pixel point outside the stroke area, as shown in FIG. The step of estimating the stroke polarity includes the following contents.

S401: In the horizontal direction, the vertical direction, and the two diagonal directions (that is, the width direction, the height direction, and the two diagonal directions of the original image), the stroke response strength is calculated by the following equation (12).

  However, w is 1/8 of the height of the original image, and f (i) indicates the luminance value of the pixel point i. By this step, four stroke response intensities in the horizontal direction, the vertical direction, and the two diagonal directions can be acquired.

  S402: The maximum stroke response intensity among the calculated four stroke response intensities is the following two conditions: [f (i) −f (l)] and [f (i) −f (k)] It is determined whether or not the two conditions that the stroke polarity is the same and the maximum stroke response intensity is larger than a predetermined threshold set in advance are satisfied. If affirmative, step S403 is executed. If not, step S404 is executed.

  In a specific application, the luminance or color values of pixel points and background pixel points inside a text stroke are generally opposite, so [f (i) −f (l)] and [f (i ) −f (k)] have the same polarity, it indicates that there is a high possibility that the i pixel point is a pixel point inside the stroke. In this step, from all the pixel points in the original image, those that may be pixel points in the stroke are detected. Here, the same polarity means that the luminance value and / or color value of the i pixel point is simultaneously larger than the l pixel point and the k pixel point, or the luminance value and / or color value of the i point are simultaneously 1 pixel. Points and smaller than k pixel points, that is, [f (i) −f (l)] and [f (i) −f (k)] are simultaneously greater than zero or less than zero. Indicates. The predetermined threshold value of the stroke response intensity can be adjusted according to actual needs. Accordingly, the present invention does not limit the selection of this predetermined threshold.

  S403: The estimated stroke polarity of the text is specified based on the polarity of [f (i) -f (l)] or [f (i) -f (k)].

  If the i pixel point satisfies the above two conditions, the estimated stroke polarity in the text is based on the polarity of [f (i) −f (l)] or [f (i) −f (k)]. Is specified as p (i), and the pixel value inside the stroke can be discriminated from the external pixel point. For example, p (i) is set to 0 when the luminance value of the pixel point inside the text stroke is lower than the luminance value of the background image point. Accordingly, p (i) of 1 indicates that the luminance value of the pixel point inside the text stroke is greater than or equal to the luminance value of the background pixel point.

  S404: The calculated stroke response strength is selected in order according to the magnitude, and step S402 is executed.

  When the maximum stroke response strength does not satisfy the above two conditions, the second largest stroke response strength is selected and the determination step of step S402 is executed to obtain the stroke response strength satisfying the above two conditions. Until this is done, analogy is performed in order according to the magnitude of the stroke response intensity. Alternatively, when none of the above four stroke response intensities satisfies the above two conditions, the pixel point i is set as a non-stroke pixel point.

  In addition, in the case where the polarity indicates the relationship of the magnitude of the luminance value between the pixel point inside the stroke area and the pixel point outside the stroke area, there are other embodiments. FIG. 5 is another flowchart showing step S302 in the second type of text enhancement method. As shown in FIG. 5, the step of estimating the stroke polarity of the text in the original image specifically includes the following contents.

  S501: In one direction, the stroke response intensity of each original pixel point in the original image is calculated by the above equation (12). This one direction is one of a horizontal direction, a vertical direction, and two diagonal directions.

  In this step, first, the stroke response strength is calculated in one of the horizontal direction, the vertical direction, and the two diagonal directions.

  S502: The stroke response strength has the following two conditions, that is, [f (i) −f (l)] and [f (i) −f (k)] have the same polarity, and the stroke response strength is preset. It is determined whether or not the predetermined threshold value is satisfied at the same time. If affirmative, step S503 is executed. When the result is negative, no process is performed on the calculated stroke response intensity, and the stroke response intensity of each original pixel point in the original image is calculated in another direction in which no calculation is performed.

  S503: The initial polarity of the original pixel point i is specified based on the polarity of [f (i) −f (l)] or [f (i) −f (k)].

  When the calculated stroke response intensity already satisfies the above two conditions, according to the polarity of [f (i) −f (l)] or [f (i) −f (k)] Specify the initial polarity. The initial polarity specified here should be the same as the polarity of [f (i) −f (l)] or [f (i) −f (k)].

  S504: It is determined whether or not the calculation of the stroke response intensities in the four directions has been completed. If affirmative, step S505 is executed. If negative, step S501 is executed.

  Next, it is determined whether or not the calculation of the stroke response strengths in the four directions of the horizontal direction, the vertical direction, and the two diagonal directions has been completed. If the calculation has not been completed, the stroke response strength is calculated. It is necessary to select one of the directions that is not present and execute step S501.

  S505: The initial polarity corresponding to the maximum stroke response intensity in the four directions is specified as the estimated stroke polarity of the text.

  When the calculation of the stroke response intensities in the four directions is completed, the initial polarity corresponding to the maximum stroke response strength is selected from the stroke response intensities satisfying the above two conditions. Specify the estimated stroke polarity. That is, the estimated stroke polarity of the text in the original image is the same as the corresponding initial polarity of the maximum stroke response intensity.

  FIG. 6 is another flowchart showing step S302 in the second type of text enhancement method. The step of estimating the stroke polarity of the text in the original image when the polarity indicates the relationship of the magnitude of the color value between the pixel point inside the stroke area and the pixel point outside the stroke area is shown in FIG. Specifically, the following contents are included.

S601: The stroke response strength of each channel is calculated by the following formula (13) in the horizontal direction, the vertical direction, and the two diagonal directions.

However, w is 1/8 of the height of the original image, and f _n (i) indicates a color value in the channel n of the pixel point i. For example, when the n channel is the R channel, the G channel, and the B channel, respectively, the stroke response strength is the sum of the stroke response strengths in the R channel, the G channel, and the B channel.

S602: Of the four stroke response intensities calculated for each channel, the maximum stroke response strength is the following two conditions, that is, [f _n (i) −f _n (l)] and [f _n ( It is determined whether or not the polarities of i) −f _n (k)] match and that the stroke response intensity is greater than a predetermined threshold value set in advance. If the result is affirmative, step S603 is executed, and if the result is negative, step S604 is executed.

In this step, it is necessary to satisfy the condition that the polarities of [f _n (i) −f _n (l)] and [f _n (i) −f _n (k)] match in any channel. .

S603: The estimated stroke polarity of the text is specified based on the polarity of [f _n (i) −f _n (l)] or [f _n (i) −f _n (k)].

  S604: The calculated stroke response intensity is selected in order according to the magnitude, and step S602 is executed.

  When the maximum stroke response strength does not satisfy the above two conditions, in this step, the second, third, and fourth stroke response strengths in order according to the size until a certain stroke response strength satisfies the above two conditions. Is selected and step S602 is executed. Alternatively, the pixel point i is set as a non-stroke pixel point when none of the four stroke response intensities satisfies the above two conditions. For example, when the second largest stroke response intensity already satisfies the above two conditions, the stroke polarity estimation flow is stopped.

  In an actual application, other embodiments exist when the polarity indicates the relationship of the magnitude of the color value between the pixel points inside the stroke area and the pixel points outside the stroke area. FIG. 7 is another flowchart showing step S302 in the second type of text enhancement method. As shown in FIG. 7, the step of estimating the stroke polarity of the text in the original image includes the following contents.

  S701: In one direction, the stroke response intensity of each channel of each original pixel point in the original image is calculated by the above equation (13), and this one direction is any one of the horizontal direction, the vertical direction, and two diagonal directions. It is.

S702: The stroke response intensity of each channel in one direction has the following two conditions, that is, [f _n (i) −f _n (l)] and [f _n (i) −f _n (k) in channel n. ] And the stroke response intensity are simultaneously satisfied that the stroke response intensity is greater than a predetermined threshold value set in advance. If affirmative, step S703 is executed. In the negative case, no initial polarity is provided in this direction.

S703: The initial polarity of the original pixel point i is specified based on the polarity of [f _n (i) −f _n (l)] or [f _n (i) −f _n (k)].

When the stroke response intensity in one direction calculated in step S701 satisfies the above two conditions, the initial polarity of the original pixel point i in this direction is set to [f _n (i) −f _n (l)] or Install in the same way as the polarity of [f _n (i) −f _n (k)].

  S704: It is determined whether or not the calculation of the stroke response intensities in the four directions has been completed. If the result is affirmative, step S705 is executed, and if the result is negative, step S701 is executed.

  S705: The initial polarity corresponding to the maximum stroke response intensity in the four directions is specified as the estimated stroke polarity of the text.

  When the calculation of the stroke response strength in all four directions is completed, select the corresponding initial polarity of the maximum stroke response strength from the stroke response strength that satisfies the above two conditions in the four directions. , The estimated stroke polarity of the above text. If the calculation of the stroke response intensities in the four directions has not been completed, any one of the directions not calculated is selected and step S701 is executed.

  Refer to FIG. 3 again.

  S303: Based on the direct difference degree and the indirect difference degree from an arbitrary original pixel point in the original image to each neighboring pixel point in the neighborhood set, the strokes of the original luminance value and / or the primary color value of each original pixel point are drawn. Dimensional filtering is performed to obtain an updated luminance value or / and an updated color value after filtering of the original image. The range of the neighborhood set is a square whose center is the original pixel point and whose side length is w, and w is smaller than the height of the original image.

  About the detail of this step, since the related content of a present Example can be referred, it abbreviate | omits here. What should be explained here is that this step may be executed simultaneously with step S302, or step S303 may be executed after step S303 is executed.

  S304: For each pixel point of the original image, it is determined whether or not the updated brightness value or / and the updated color value after filtering are matched with the above-described stroke polarity. If the result is affirmative, step S305 is executed and negative In this case, no replacement is performed.

  After the stroke polarity is obtained by calculation and the stroke filtering is performed, it is sequentially determined whether or not the updated brightness value or / and the updated color value after filtering of each pixel point of the original image matches the stroke polarity described above. be able to. If the updated luminance value or / and updated color value after filtering of a certain pixel point does not match the above-mentioned stroke polarity, the original luminance value or / and color value of this pixel point is not replaced, and the next pixel point is filtered. It continues to determine whether or not the later updated luminance value or / and updated color value match the above-described stroke polarity.

  FIG. 8 is a flowchart showing step S304 in the second type of text enhancement method. As shown in FIG. 8, the step S304 specifically includes the following contents.

  S801: A first magnitude relationship between an updated luminance value or / and updated color value after filtering and a primary luminance value or / and primary color value is acquired.

  First, a first magnitude relationship between the updated luminance value or / and the updated color value after filtering and the primary luminance value or / and the primary color value is acquired. Regarding the relationship of the first magnitude here, for example, when the updated luminance value is brighter than the luminance of the original luminance value, the first magnitude relationship is that the updated luminance value is larger than the original luminance value, or For example, when the updated color value is larger than the primary color value, the relationship of the first magnitude is that the updated color value is larger than the primary color value. This first magnitude relationship can be obtained by performing algebraic subtraction on the updated luminance value or / and updated color value and the primary luminance value or / and primary color value.

  S802: It is determined whether or not the first magnitude relationship matches the second magnitude relationship indicated by the above-described stroke polarity.

  Since the above-mentioned stroke polarity indicates the relationship between the luminance value or / and color value of the pixel point inside the stroke and the luminance value or / and color value of the pixel point outside the stroke, the updated color value is greater than the primary color value. If the updated color value is greater than the primary color value, or if the updated color value is less than the primary color value and the second magnitude indicated by the stroke polarity When the updated color value is smaller than the primary color value, the relationship between the first size and the relationship between the second size shown in the above-described stroke polarity is considered to be matched. Conversely, it is considered that the relationship between the first size and the relationship between the second size do not match.

  Refer to FIG. 3 again.

  S305: Replace the corresponding original luminance value or / and primary color value with the updated luminance value or / and updated color value after filtering, respectively, to generate a text-enhanced image corresponding to the original image.

  When the relationship of the first magnitude and the relation of the second magnitude are combined, the corresponding primary luminance value or / and primary color value is replaced with the updated luminance value or / and updated color value after filtering, respectively. By doing so, the text emphasis image corresponding to the original image can be acquired. In this embodiment, when the stroke polarity is combined with the updated luminance value or the updated color value after filtering in the stroke polarity estimation step, the text enhancement can be further performed. As compared with the above, the effect of text emphasis becomes more prominent and can contribute to the improvement of the accuracy of subsequent text extraction.

  According to the above-described text emphasizing method according to the embodiment of the present invention, the text stroke at the pixel point in the acquired text-enhanced image is enhanced, the matching of the pixels inside the stroke is enhanced, and the difference between the text and the background The text was deepened, ie the text contained in the original image was emphasized. This makes subsequent text extraction for this text-enhanced image more accurate and accurate.

  In addition, an embodiment of the present invention provides a first type of text enhancement device corresponding to the above-described first type of text enhancement method. FIG. 9 is a schematic diagram showing a first type of text emphasizing apparatus according to an embodiment of the present invention.

  As shown in FIG. 9, the first type of text enhancement device can include an acquisition module 901 for acquiring an original image including at least one line of text.

  Furthermore, as shown in FIG. 9, the first type of text enhancement device is based on the direct difference degree and the indirect difference degree from an arbitrary original pixel point in the original image to each neighboring pixel point in the neighborhood set. A filter module 902 is provided for performing two-dimensional handwriting filtering on the original luminance value or / and primary color value of each original pixel point to obtain an updated luminance value or / and updated color value after filtering of the original image. Can do. The range of the neighborhood set is a square centered on the original pixel point and the side length is w, and w is smaller than the height of the original image.

  FIG. 10 is a schematic diagram showing the filter module 902 in the first type of text enhancement device.

As shown in FIG. 10, the filter module 902
A first acquisition sub-module 1001 for performing algebraic subtraction on the original luminance value or / and the primary color value of the original pixel point and each neighboring pixel point to directly acquire the difference degree;
A second acquisition sub-module 1002 for acquiring an indirect difference based on a gradient module from the original pixel point to each neighboring pixel point in the neighborhood set;
A weight calculation submodule 1003 for calculating a luminance value and / or a color value weight value for the original pixel point of each neighboring pixel point based on the direct difference degree and the indirect difference degree;
An updated luminance value calculation submodule 1004 for calculating an updated luminance value of the original pixel point by the above-described stroke two-dimensional filter equation (10);
An updated color value calculation submodule 1005 for calculating the updated color value of the original pixel point can be provided by the above-described stroke two-dimensional filter equation (11).

  Refer to FIG. 9 again. The first type of text enhancement device generates a text enhanced image corresponding to the original image by replacing the corresponding original luminance value or / and color value with the updated luminance value or / and updated color value after filtering, respectively. A replacement module 903 can be provided.

  According to the above text emphasis device according to the embodiment of the present invention, after obtaining the updated luminance value or / and the updated color value after filtering, the updated luminance value or / and the updated color value are used as the primary luminance value or / and the primary color. By substituting each value, the text stroke at the pixel point in the original image is emphasized, the pixel matching inside the stroke is enhanced, and the difference between the text and the background is deepened, so that the subsequent text extraction Therefore, it is possible to provide a better text-enhanced image and improve the accuracy and precision of subsequent text extraction.

  In addition, an embodiment of the present invention further provides a second type of text enhancement device corresponding to the above-described second type of text enhancement method. FIG. 11 is a schematic diagram showing a second type of text emphasizing apparatus according to an embodiment of the present invention.

  As shown in FIG. 11, the second type of text enhancement device includes an acquisition module 901 for acquiring an original image including at least one line of text, and a stroke polarity for estimating the stroke polarity of the text in the original image. An estimation module 1101. Here, the polarity indicates the relationship between the magnitude of the luminance value and / or the color value between the pixel point located inside the stroke area and the pixel point located outside the stroke area.

  The stroke polarity estimation module 1101 can perform different specific arrangements in different application scenes. FIG. 12 is a schematic diagram showing the stroke polarity estimation module 1101 in the second type of text enhancement device.

In a certain scene, when the polarity indicates the relationship of the magnitude of the luminance value between the pixel point inside the stroke area and the pixel point outside the stroke area, as shown in FIG. 12, the stroke polarity estimation module 1101 Is
A first calculation sub-module 1201 for calculating the stroke response intensity according to the above equation (12) in the horizontal direction, the vertical direction, and the two diagonal directions;
Of the four calculated stroke response intensities, the maximum stroke response intensity has the following two conditions, that is, [f (i) −f (l)] and [f (i) −f (k)] have the same polarity And a first determination submodule 1202 for determining whether or not the stroke response strength satisfies a predetermined threshold value larger than a preset threshold value;
When the determination result by the first determination submodule 1202 is affirmative, the estimated stroke polarity of the text is determined based on the polarities of [f (i) −f (l)] and [f (i) −f (k)]. A first identification sub-module 1203 for identifying
When the determination result by the first determination submodule 1202 is negative, the calculated stroke response strength is selected in order according to the magnitude until a certain stroke response strength satisfies the above two conditions, and the first determination submodule A first trigger sub-module 1204 for setting the pixel point i as a non-stroke pixel point when none of the four stroke response intensities satisfies the two conditions, Can be provided.

  FIG. 13 is another schematic diagram showing the stroke polarity estimation module 1101 in the second type of text enhancement device.

In other scenes, when the polarity indicates the relationship of the magnitude of the luminance value between the pixel point inside the stroke area and the pixel point outside the stroke area, the stroke polarity estimation module 1101 is shown in FIG. ,
A second calculation sub-module 1301 for calculating the stroke response strength of each original pixel point in the original image according to the above equation (12) in any of the horizontal direction, the vertical direction, and the two diagonal directions;
The stroke response strength is the following two conditions, that is, [f (i) −f (l)] and [f (i) −f (k)] have the same polarity and the stroke response strength is set in advance. A second determination sub-module 1302 for determining whether or not simultaneously satisfying that a value greater than a threshold value of
If the determination result by the second determination submodule 1302 is affirmative, the original pixel point i is determined based on the polarities of [f (i) −f (l)] and [f (i) −f (k)]. A second identification sub-module 1303 for identifying the initial polarity;
A third determination submodule 1304 for determining whether or not the calculation of the stroke response intensities in the four directions has been completed;
A third identification sub-module 1305 for identifying the corresponding initial polarity of the maximum stroke response intensity in the four directions as the estimated stroke polarity of the text when the determination result by the third determination sub-module 1304 is positive; ,
And a second trigger submodule 1306 for triggering (activating) the second calculation submodule when the determination result by the third determination submodule 1304 is negative.

  FIG. 14 is another schematic diagram showing the stroke polarity estimation module 1101 in the second type of text enhancement device.

In other scenes, when the polarity indicates the relationship of the magnitude of the color value between the pixel point inside the stroke area and the pixel point outside the stroke area, as shown in FIG. Module 1101
A third calculation sub-module 1401 for calculating the stroke response intensity according to the above equation (13) in the horizontal direction, the vertical direction, and the two diagonal directions;
Among the calculated stroke response intensities, the maximum stroke response intensity has the following two conditions, namely [f _n (i) −f _n (l)] and [f _n (i) −f _n (k) in channel n. And the fourth response sub-module 1402 for determining whether or not the stroke response intensity is greater than a predetermined threshold value,
If the determination result by the fourth determination submodule 1402 is affirmative, the text is determined based on the polarity of [f _n (i) −f _n (l)] or [f _n (i) −f _n (k)]. A fourth identification sub-module 1403 for identifying the estimated stroke polarity of
If the determination result by the fourth determination submodule 1402 is negative, the stroke response strength calculated in order according to the magnitude relationship is selected until a certain stroke response strength satisfies the above two conditions, and the fourth determination A third trigger sub-module for triggering (activating) the sub-module 1402 or making the pixel point i a non-stroke pixel point when none of the four stroke response intensities satisfy the above two conditions 1404.

  FIG. 15 is another schematic diagram showing the stroke polarity estimation module 1101 in the second type of text enhancement device.

In other scenes, when the polarity indicates the relationship of the magnitude of the color value between the pixel point inside the stroke area and the pixel point outside the stroke area, as shown in FIG. 15, the stroke polarity estimation module 1101 is
A fourth calculation submodule 1501 for calculating the stroke response strength of each original pixel point in the original image according to the above equation (13) in any one of the horizontal direction, the vertical direction, and the two diagonal directions;
The stroke response intensity is the following two conditions, that is, in channel n, [f _n (i) −f _n (l)] and [f _n (i) −f _n (k)] have the same polarity, and this A fifth determination sub-module 1502 for determining whether or not the stroke response intensity is simultaneously greater than a predetermined threshold value,
If the determination result by the fifth determination submodule 1502 is positive, based on the polarity of [f _n (i) −f _n (l)] or [f _n (i) −f _n (k)] A fifth identification sub-module 1503 for identifying the initial polarity of the pixel point i;
A sixth determination submodule 1504 for determining whether or not the calculation of the stroke response intensities in the four directions has been completed;
A sixth identification submodule 1505 for identifying the corresponding initial polarity of the maximum stroke response intensity in the four directions as the estimated stroke polarity of the text when the determination result by the sixth determination submodule 1504 is positive; ,
And a fourth trigger submodule 1506 for triggering (activating) the fourth calculation submodule 1501 when the determination result by the sixth determination submodule 1505 is negative.

  Refer to FIG. 11 again. The second type of text emphasizing device further provides an original luminance value of each original pixel point based on a direct difference degree and an indirect difference degree from an arbitrary original pixel point in the original image to each neighboring pixel point in the neighborhood set. Alternatively, a filter module 902 may be included for performing stroke two-dimensional filtering on the primary color values to obtain updated luminance values or / and updated color values after filtering of the original image. The range of the neighborhood set is a square whose center is the original pixel point and whose side length is w, and the w is smaller than the height of the original image.

  Further, as shown in FIG. 11, the second type of text enhancement device further determines whether or not the updated luminance value after filtering and / or the updated color value and the stroke polarity are matched. A determination module 1102 for triggering (activating) the replacement module 903 can be provided.

  FIG. 16 is a schematic diagram showing the determination module 1102 in the second type of text enhancement device.

As shown in FIG. 16, in an actual application, the decision module 1102
A third acquisition submodule 1601 for acquiring a first magnitude relationship between the updated luminance value or / and the updated color value after filtering and the primary luminance value or / and the primary color value;
A seventh determination submodule 1602 for determining whether or not the first magnitude relationship and the second magnitude relationship indicated by the stroke polarity are matched can be provided.

  Since the relationship between the first size and the relationship between the second size has already been described in the description related to FIG. 8 showing the flow chat in step S304 in the above-described second type of text enhancement method, Omitted.

  Refer to FIG. 11 again. The second type of text emphasizing device further replaces the corresponding original luminance value or / and primary color value with the updated luminance value or / and updated color value after filtering, respectively, so that a text enhanced image corresponding to the original image is obtained. A replacement module 903 for generating may be provided.

  According to the above-described apparatus according to the embodiment of the present invention, the updated color value or the updated luminance value after filtering can be verified by adopting the form of the stroke polarity estimation. When the stroke polarity is combined with the updated color value or / and the updated luminance value after filtering, the corresponding primary luminance value or / and the primary color value are respectively replaced with the updated luminance value or / and the updated color value after filtering. As a result, the acquired text-enhanced image becomes more effective and accurate.

  FIG. 17 is a flowchart illustrating a text extraction method according to an embodiment of the present invention. As shown in FIG. 17, one embodiment of the present invention further provides a text extraction method after text enhancement. This text extraction method includes the following steps. Note that this text extraction method may be executed by a computer, for example.

  S1701: An original image including at least one line of text is acquired.

  S1702: Based on the direct difference degree and the indirect difference degree from an arbitrary original pixel point in the original image to each neighboring pixel point in the neighborhood set, a stroke is drawn for the original luminance value or / and the primary color value of each original pixel point. Two-dimensional filtering is performed to obtain an updated luminance value or / and an updated color value after filtering of the original image. The range of the neighborhood set is a square whose center is the original pixel point and whose side length is w, and w is smaller than the height of the original image.

  S1703: The text-enhanced image corresponding to the original image is generated by replacing the corresponding original luminance value or / and primary color value with the updated luminance value or / and updated color value after filtering, respectively.

  S1704: Extract the text in the text enhanced image.

  If the above-described text extraction method is employed, text extraction can be performed based on the text enhanced image. The text extracted in this way becomes more accurate and accurate. At the same time, since the text is already emphasized at the time of extraction, the complexity of text extraction can be reduced and the efficiency of text extraction can be improved.

FIG. 18 is a schematic diagram illustrating a text extraction apparatus according to an embodiment of the present invention. As shown in FIG. 18, the embodiment of the present invention further provides a text extraction apparatus corresponding to the text extraction method described above. This text extraction device
An acquisition module 901 for acquiring an original image including at least one line of text;
Based on the direct difference and indirect difference from an arbitrary original pixel point in the original image to each neighboring pixel point in the neighborhood set, the two-dimensional strokes for the original luminance value and / or the primary color value of each original pixel point A filter module 902 for performing filtering to obtain an updated luminance value or / and an updated color value after filtering of the original image, wherein the neighborhood set range is centered on the original pixel point and the side length is w And w is a filter module 902 smaller than the height of the original image,
A replacement module 903 for generating a text-enhanced image corresponding to the original image by replacing the corresponding original luminance value or / and color value with the updated luminance value or / and the updated color value after filtering, respectively;
An extraction module 1801 for extracting text in the text-enhanced image.

  When the above-described text extraction device is employed, text extraction can be performed based on the text enhanced image. The text extracted in this way becomes more accurate and accurate. At the same time, since the text is already emphasized during extraction, the complexity of text extraction can be reduced and the efficiency of text extraction can be improved.

  Further, it should be described here that the above-described series of processes or apparatuses may be realized by software and / or firmware. When implemented by software and / or firmware, various programs that configure the software and / or firmware from a storage medium or network to a computer having a dedicated hardware structure, such as the general-purpose computer 1900 shown in FIG. By installing, this computer can execute various functions.

  FIG. 19 is a block diagram showing a schematic structure of a computer as an information processing apparatus according to an embodiment of the present invention. In FIG. 19, the central processing unit (CPU) 1901 has various programs based on programs stored in a read-only memory (ROM) 1902 or various programs loaded from a storage unit 1908 to a random access memory (RAM) 1903. Execute the process. The RAM 1903 also stores data necessary when the CPU 1901 executes various processes as necessary.

  The CPU 1901, ROM 1902 and RAM 1903 are connected to each other via a bus 1904. An input / output interface 1905 is also connected to the bus 1904.

  An input unit 1906 including a keyboard and a mouse, an output unit 1907 including a display and a speaker such as a cathode ray tube (CRT) or a liquid crystal display (LCD), a storage unit 1908 including a hard disk, etc. A communication unit 1909 including a network interface card such as a card and a modem is connected to an input / output interface 1905. A communication unit 1909 executes communication processing via a network such as the Internet.

  A drive 1910 is also connected to the input / output interface 1905 as needed. A removable medium 1911 such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory is attached to the drive 1910 as necessary, and a computer program read out from the drive 1910 is stored in the storage unit 1908 as necessary. To be installed.

  When the above-described series of processing is realized by software, a program constituting the software is installed from a network (for example, the Internet) or a storage medium (for example, removable medium 1911).

  Those skilled in the art will recognize that such storage media is not limited to the removable media 1911 shown in FIG. 19 in which the program is stored and delivered remotely from the device to provide the program to the user. Should be understood. Examples of the removable medium 1911 include a magnetic disk (including a floppy disk (registered trademark)), an optical disk (including a compact disk read-only memory (CD-ROM) and a digital versatile disk (DVD)), and a magneto-optical disk ( Including mini disk (MD) (registered trademark) and semiconductor memory. Alternatively, the storage medium may be a ROM 1902 or a hard disk included in the storage unit 1908. They store programs and are distributed to users along with the devices that contain them.

  Further, it should be explained that the above-described series of processing steps can be executed in the order of time as described, but not necessarily in the order of time. Some steps can be performed in parallel or independently of each other.

  The preferred embodiment of the present invention has been described above, but the present invention is not limited to this embodiment, and all modifications to the present invention belong to the technical scope of the present invention unless departing from the spirit of the present invention.

  The following additional notes are further disclosed with respect to the embodiment including the above examples.

  (Supplementary Note 1) A method in which a computer emphasizes text in an image, wherein the computer acquires an original image including at least one line of text, and an arbitrary original pixel point in the original image and its vicinity Based on the direct difference and the indirect difference to each neighboring pixel point in the set, the two-dimensional filtering is performed on the original luminance value or / and the primary color value of each original pixel point, and after the original image is filtered An update value acquisition step of acquiring an update brightness value or / and an update color value, and replacing the corresponding primary brightness value or / and primary color value with the updated brightness value or / and update color value after filtering, Executing an emphasized image generation step of generating a text-enhanced image corresponding to the original image, and the range of the neighborhood set is centered on the original pixel point And, and side length becomes square is w, w is less than the height of the original image method.

ここで、N(i)は画素点iの近傍集合を示し、w(i,j)は前記近傍集合内の近傍画素点ｊの、画素点iに対する輝度値の重み値を示し、f(j)は前記近傍画素点jの原輝度値であるステップ、及び／又は、以下の筆画二次元フィルタ式を採用して前記原画素点の更新色値を算出するステップであって、

ここで、w_n(i,j)はｎチャネルにおいて前記近傍画素点jの、原画素点iに対する色値の重み値を示し、f_n(j)はｎチャネルにおいて前記近傍画素点ｊの原色値であるステップを含む、付記１に記載の方法。 (Additional remark 2) The said update value acquisition step performs the algebraic subtraction with respect to the primary luminance value or / and primary color value of the said original pixel point and each neighboring pixel point, and acquires the said direct difference degree, The said original pixel Obtaining the indirect difference based on a gradient module from a point to each neighboring pixel point of the neighborhood set, a luminance value of each neighboring pixel point with respect to the original pixel point based on the direct difference and the indirect difference Or / and a step of calculating a weight value of the color value, a step of calculating an updated luminance value of the original pixel point by employing the following stroke two-dimensional filter equation,

Here, N (i) represents a neighborhood set of the pixel point i, w (i, j) represents a weight value of the luminance value of the neighborhood pixel point j in the neighborhood set with respect to the pixel point i, and f (j ) Is an original luminance value of the neighboring pixel point j, and / or a step of calculating an updated color value of the original pixel point using the following stroke two-dimensional filter equation,

Here, w _n (i, j) represents the weight value of the color value of the neighboring pixel point j with respect to the original pixel point i in the n channel, and f _n (j) represents the primary color of the neighboring pixel point j in the n channel. The method of claim 1, comprising the step of being a value.

  (Additional remark 3) The said computer is a polarity estimation step which estimates the stroke polarity of the text in the said original image after the said original image acquisition step, Comprising: The said polarity is a pixel point located inside a stroke area, and a stroke Polarity estimation step indicating the relationship between the magnitude of the primary luminance value and / or the primary color value between pixel points located outside the region, and the updated luminance value after filtering or / and after the updated value acquisition step The method according to appendix 1, wherein it is determined whether or not the updated color value and the stroke polarity are matched, and if the result is affirmative, an updated value determining step of replacing the primary luminance value or / and the primary color value is executed.

ここで、wは前記原画像の高さの８分の一であり、f(i)は画素点iの輝度値を示すステップ、及び、算出した四つの筆画応答強度のうち最大の筆画応答強度が次の二つの条件、即ち、[f(i)−f(l)]と[f(i)−f(k)]の極性が同じ、且つ前記最大の筆画応答強度が所定の閾値よりも大きいという二つの条件を満たしているか否かを判断し、肯定の場合に、[f(i)−f(l)]と[f(i)−f(k)]の極性に基づいて前記テキストの推定筆画極性を特定し、否定の場合に、ある筆画応答強度が前記二つの条件を満たすまで、算出された前記筆画応答強度を大きさに従って順に選択して前記判断ステップを実行し、或いは、前記四つの筆画応答強度がいずれも前記二つの条件を満たさない場合に、前記画素点iを非筆画画素点とするステップを含む、付記３に記載の方法。 (Supplementary Note 4) When the polarity indicates the relationship of the magnitude of the luminance value between the pixel point inside the stroke area and the pixel point outside the stroke area, the polarity estimation step includes the horizontal direction, the vertical direction, and the two In the two diagonal directions, the following formula is used to calculate the stroke response strength, respectively,

Here, w is one-eighth of the height of the original image, f (i) is a step indicating the luminance value of the pixel point i, and the maximum stroke response strength among the four stroke response strengths calculated. Are the following two conditions, that is, [f (i) −f (l)] and [f (i) −f (k)] have the same polarity and the maximum stroke response intensity is lower than a predetermined threshold value. It is determined whether or not two conditions of large are satisfied, and in the case of affirmation, the text is based on the polarities of [f (i) −f (l)] and [f (i) −f (k)]. The estimated stroke polarity is identified, and in the case of negative, until the certain stroke response strength satisfies the two conditions, the calculated stroke response strength is sequentially selected according to the magnitude, and the determination step is executed, or The method according to appendix 3, comprising the step of setting the pixel point i as a non-stroke pixel point when none of the four stroke response intensities satisfy the two conditions.

ここで、wは前記原画像の高さの８分の一であり、f(i)は画素点iの輝度値を示すステップ、前記筆画応答強度が次の二つの条件、即ち、[f(i)−f(l)]と[f(i)−f(k)]の極性が同じ、且つ前記筆画応答強度が所定の閾値より大きいという二つの条件を同時に満たすか否かを判断し、肯定の場合に、[f(i)−f(l)]又は[f(i)−f(k)]の極性に基づいて前記原画素点iの初期極性を特定するステップ、及び、四つの方向上の筆画応答強度の算出がすべて完了したか否かを判断し、肯定の場合に、四つの方向上の最大の筆画応答強度が対応する初期極性を、前記テキストの推定筆画極性として特定し、否定の場合に、前記筆画応答強度の算出ステップを繰り返すステップを含む、付記３に記載の方法。 (Supplementary Note 5) In the case where the polarity indicates the relationship of the magnitude of the luminance value between the pixel point inside the stroke area and the pixel point outside the stroke area, the polarity estimation step includes the horizontal direction, the vertical direction, and the two In any one of the two diagonal directions, the following formula is adopted to calculate the stroke response strength of each original pixel point in the original image,

Here, w is one-eighth of the height of the original image, f (i) is a step indicating the luminance value of the pixel point i, and the stroke response intensity is the following two conditions: [f ( i) −f (l)] and [f (i) −f (k)] have the same polarity and determine whether or not the two conditions that the stroke response intensity is greater than a predetermined threshold are satisfied simultaneously, If yes, identifying the initial polarity of the original pixel point i based on the polarity of [f (i) −f (l)] or [f (i) −f (k)], and It is determined whether or not the calculation of the stroke response intensity in the direction has been completed, and in the case of affirmation, the initial polarity corresponding to the maximum stroke response intensity in the four directions is specified as the estimated stroke polarity of the text. The method according to supplementary note 3, including a step of repeating the step of calculating the stroke response strength in the case of negative.

ここで、wは前記原画像の高さの８分の一であり、f_n(i)は画素点iのチャネルｎにおける色値を示すステップ、算出した四つの筆画応答強度のうち最大の筆画応答強度が次の二つの条件、即ち、チャネルｎにおいて[f_n(i)−f_n(l)]と[f_n(i)−f_n(l)]との極性が一致し、且つ前記筆画応答強度が所定の閾値より大きいという二つの条件を満たしているか否かを判断し、肯定の場合に、[f_n(i)−f_n(l)]又は[f_n(i)−f_n(k)]の極性に基づいて前記テキストの推定筆画極性を特定し、否定の場合に、ある筆画応答強度が前記二つの条件を満たすまで、算出した前記筆画応答強度を大きさに従って順に選択して前記判断ステップを実行し、或いは、前記四つの筆画応答強度がいずれも前記二つの条件を満たしていない場合に、前記画素点iを非筆画画素点とするステップを含む、付記３に記載の方法。 (Supplementary note 6) When the polarity indicates the relationship of the magnitude of the color value between the pixel point inside the stroke area and the pixel point outside the stroke area, the polarity estimation step includes the horizontal direction, the vertical direction, and two In the diagonal direction, the following formula is used to calculate the stroke response strength, respectively,

Here, w is one-eighth of the height of the original image, f _n (i) is a step indicating the color value in the channel n at the pixel point i, and the largest stroke among the calculated four stroke response intensities. The response intensity has the following two conditions, that is, the polarities of [f _n (i) −f _n (l)] and [f _n (i) −f _n (l)] match in channel n, and It is determined whether or not two conditions that the stroke response intensity is larger than a predetermined threshold are satisfied, and if affirmative, [f _n (i) −f _n (l)] or [f _n (i) −f _n (k)] specifies the estimated stroke polarity of the text, and in the negative case, the calculated stroke response strength is selected in order according to the magnitude until a certain stroke response strength satisfies the two conditions. The determination step is executed, or when none of the four stroke response intensities satisfy the two conditions, the pixel point i is set as a non-stroke pixel point. Including steps, method of statement 3.

ここで、wは前記原画像の高さの８分の一であり、f_n(i)は画素点iのチャネルｎにおいる色値を示すステップ、前記筆画応答強度が次の二つの条件、即ち、チャネルｎにおいて[f_n(i)−f_n(l)]と[f_n(i)−f_n(k)]との極性が一致し、且つ前記筆画応答強度が所定の閾値より大きいという二つの条件を同時に満たしているか否かを判断し、肯定の場合に、[f_n(i)−f_n(l)]又は[f_n(i)−f_n(k)]の極性に基づいて前記原画素点iの初期極性を特定するステップ、及び、四つの方向上の筆画応答強度の算出がすべて完了したか否かを判断し、肯定の場合に、四つの方向上の最大の筆画応答強度が対応する初期極性を、前記テキストの推定筆画極性として特定し、否定の場合に、前記筆画応答強度算出ステップを繰り返すステップを含む、付記３に記載の方法。 (Supplementary note 7) In the case where the polarity indicates the relationship of the magnitude of the color value between the pixel point inside the stroke area and the pixel point outside the stroke area, the polarity estimation step includes the horizontal direction, the vertical direction, and the two In any one of the two diagonal directions, the following formula is adopted to calculate the stroke response strength of each original pixel point in the original image,

Here, w is one-eighth of the height of the original image, f _n (i) is a step indicating the color value in the channel n at the pixel point i, and the stroke response intensity is the following two conditions: That is, in channel n, [f _n (i) −f _n (l)] and [f _n (i) −f _n (k)] have the same polarity, and the stroke response intensity is greater than a predetermined threshold value. Judgment is made whether the two conditions of large are satisfied at the same time. If the result is affirmative, the polarity of [f _n (i) −f _n (l)] or [f _n (i) −f _n (k)] Determining the initial polarity of the original pixel point i based on the above, and determining whether or not the calculation of the stroke response intensity in the four directions has been completed. The method according to claim 3, further comprising: specifying the initial polarity corresponding to the stroke response intensity of the text as the estimated stroke polarity of the text, and repeating the stroke response intensity calculating step when negative.

  (Supplementary Note 8) The update value determination step includes obtaining a first magnitude relationship between the updated luminance value or / and the updated color value after filtering and the primary luminance value or / and the primary color value; and The method according to claim 3, further comprising the step of determining whether or not the first magnitude relationship matches the second magnitude relationship indicated by the stroke polarity.

  (Supplementary note 9) An apparatus for emphasizing text in an image, an acquisition module that acquires an original image including at least one line of text, and an arbitrary original pixel point in the original image to each neighboring pixel point in the neighborhood set Based on the direct difference degree and the indirect difference degree, the original luminance value or / and primary color value of each original pixel point is subjected to stroke two-dimensional filtering, and the updated luminance value or / and updating after filtering of the original image is performed. A text enhancement image corresponding to the original image by replacing the corresponding original luminance value or / and primary color value with the filter module for obtaining a color value and the updated luminance value or / and the updated color value after filtering, respectively. A range of the neighborhood set is a square centered on the original pixel point and having a side length of w, where w is the original image Height less than the device.

以下の筆画二次元フィルタ式を採用して、前記原画素点の更新色値を算出する更新色値算出サブモジュールを含み、

ここで、N(i)は画素点iの近傍集合を示し、w(i,j)は近傍画素点ｊの、原画素点iに対する輝度値の重み値を示し、f(j)は前記近傍画素点jの輝度値であり、w_n(i,j)はｎチャネルにおいて近傍画素点jの、原画素点iに対する色値の重み値を示し、f_n(j)はｎチャネルにおいて前記近傍集合内の画素点ｊの色値である、付記９に記載の装置。 (Additional remark 10) The said filter module performs the algebraic subtraction with respect to the primary luminance value or / and primary color value of the said original pixel point and each neighboring pixel point, and acquires the said direct difference degree. A module, a second acquisition sub-module that acquires the indirect difference based on a gradient module from the original pixel point to each neighboring pixel point in the neighborhood set, and each neighboring pixel based on the direct difference and the indirect difference A weight calculation submodule that calculates a luminance value or / and a color value weight value of the point with respect to the original pixel point, and an update that calculates an updated luminance value of the original pixel point by employing the following stroke two-dimensional filter equation Luminance value calculation sub-module and / or

Adopting the following stroke two-dimensional filter equation, including an updated color value calculation submodule for calculating an updated color value of the original pixel point,

Here, N (i) represents the neighborhood set of pixel point i, w (i, j) represents the weight value of the luminance value of the neighboring pixel point j with respect to the original pixel point i, and f (j) is the neighborhood The luminance value of the pixel point j, w _n (i, j) represents the weight value of the color value of the neighboring pixel point j with respect to the original pixel point i in the n channel, and f _n (j) represents the neighborhood in the n channel The apparatus of claim 9 which is the color value of pixel point j in the set.

  (Supplementary Note 11) It is determined whether or not the stroke polarity estimation module for estimating the stroke polarity of the text in the original image, the updated luminance value or / and the updated color value after filtering, and the stroke polarity are matched. A determination module for triggering the replacement module, wherein the polarity is a primary luminance value or / and a primary color between a pixel point located inside the stroke area and a pixel point located outside the stroke area The apparatus according to appendix 9, which shows a magnitude relationship.

ここで、wは前記原画像の高さの８分の一であり、f(i)は画素点iの輝度値を示す第１の算出サブモジュール、算出した四つの筆画応答強度のうち最大の筆画応答強度が次の二つの条件、即ち、[f(i)−f(l)]と[f(i)−f(k)]の極性が同じ、且つ前記筆画応答強度が所定の閾値よりも大きいという二つの条件を満たしているか否かを判断する第１の判断サブモジュール、及び、前記第１の判断サブモジュールによる結果が肯定の場合に、[f(i)−f(l)]と[f(i)−f(k)]の極性に基づいて前記テキストの推定筆画極性を特定する第１の特定サブモジュールと、前記第１の判断サブモジュールによる結果が否定の場合に、ある筆画応答強度が前記二つの条件を満たすまで、算出された前記筆画応答強度を大きさに従って順に選択して前記第１の判断サブモジュールをトリガし、或いは、前記四つの筆画応答強度がいずれも前記二つの条件を満たさない場合に、前記画素点iを非筆画画素点とする第１のトリガサブモジュールを含む、付記１１に記載の装置。 (Supplementary note 12) When the polarity indicates the relationship of the magnitude of the luminance value between the pixel point inside the stroke area and the pixel point outside the stroke area, the stroke polarity estimation module performs the horizontal direction, the vertical direction, and the two A first calculation sub-module that calculates the stroke response strength using the following formula in each of the diagonal directions,

Here, w is one-eighth of the height of the original image, f (i) is a first calculation submodule indicating the luminance value of the pixel point i, and is the largest of the calculated four stroke response intensities. The stroke response strength is the following two conditions, that is, the polarities of [f (i) −f (l)] and [f (i) −f (k)] are the same, and the stroke response strength exceeds a predetermined threshold value. [F (i) −f (l)] when the result of the first determination submodule that determines whether or not the two conditions are satisfied and the result of the first determination submodule is affirmative And [f (i) −f (k)] based on the polarity of the first specific submodule for specifying the estimated stroke polarity of the text and the result of the first determination submodule is negative Until the stroke response strength satisfies the above two conditions, the calculated stroke response strengths are selected in order according to the magnitude, and the first determination submodule is set as a trigram. Or the first trigger sub-module that uses the pixel point i as a non-stroke pixel point when none of the four stroke response intensities satisfy the two conditions. .

ここで、wは前記原画像の高さの８分の一であり、f(i)は画素点iの輝度値を示す第２の算出サブモジュール、前記筆画応答強度が次の二つの条件、即ち、[f(i)−f(l)]と[f(i)−f(k)の極性が同じ、且つ前記筆画応答強度が所定の閾値よりも大きいという二つの条件を同時に満たすか否かを判断する第２の判断サブモジュール、前記第２の判断サブモジュールによる結果が肯定の場合に、[f(i)−f(l)]又は[f(i)−f(k)]の極性に基づいて前記原画素点iの初期極性を特定する第２の特定サブモジュール、四つの方向上の筆画応答強度の算出がすべて完了したか否かを判断する第３の判断サブモジュール、前記第３の判断サブモジュールによる結果が肯定の場合に、四つの方向上の最大の筆画応答強度が対応する初期極性を、前記テキストの推定筆画極性として特定するための第３の特定サブモジュール、前記第３の判断サブモジュールによる結果が否定の場合に、前記の第２の算出サブモジュールをトリガ（起動）する第２のトリガサブモジュールを含む、付記１１に記載の装置。 (Supplementary note 13) When the polarity indicates the relationship of the magnitude of the luminance value between the pixel point inside the stroke area and the pixel point outside the stroke area, the stroke polarity estimation module is A second calculation submodule for calculating the stroke response strength of each original pixel point in the original image by adopting the following formula in any one of two diagonal directions,

Here, w is one-eighth of the height of the original image, f (i) is a second calculation submodule indicating the luminance value of the pixel point i, and the stroke response intensity is the following two conditions: That is, whether or not the two conditions that [f (i) −f (l)] and [f (i) −f (k) have the same polarity and the stroke response intensity is larger than a predetermined threshold are satisfied simultaneously. A second determination sub-module for determining whether the result of the second determination sub-module is affirmative, [f (i) −f (l)] or [f (i) −f (k)] A second specifying submodule for specifying the initial polarity of the original pixel point i based on the polarity; a third determining submodule for determining whether or not the calculation of the stroke response intensities in the four directions has been completed; If the result of the third decision submodule is positive, the initial polarity corresponding to the maximum stroke response intensity in the four directions is identified as the estimated stroke polarity of the text. A third specific submodule for determining, and a second trigger submodule that triggers (activates) the second calculation submodule when the result of the third determination submodule is negative 11. The apparatus according to 11.

ここで、wは前記原画像の高さの８分の一であり、f_n(i)はチャネルｎにおいて画素点iの色値を示す第３の算出サブモジュール、算出した四つの筆画応答強度のうち最大の筆画応答強度が次の二つの条件、即ち、チャネルｎにおいて[f_n(i)−f_n(l)]と[f_n(i)−f_n(l)]との極性が一致し、且つ前記筆画応答強度が所定の閾値よりも大きいという二つの条件を満たしているか否かを判断する第４の判断サブモジュール、前記第４の判断サブモジュールによる結果が肯定の場合に、[f_n(i)−f_n(l)]又は[f_n(i)−f_n(k)]の極性に基づいて前記テキストの推定筆画極性を特定する第４の特定サブモジュール、及び、前記第４の判断サブモジュールによる結果が否定の場合に、ある筆画応答強度が前記二つの条件を満たすまで、算出した前記筆画応答強度を大きさに従って順に選択して前記第４の判断サブモジュールをトリガし、或いは、前記四つの筆画応答強度がいずれも前記二つの条件を満たしていない場合に、前記画素点iを非筆画画素点とする第３のトリガサブモジュールを含む、付記１１に記載の装置。 (Supplementary note 14) When the polarity indicates the relationship of the magnitude of the color value between the pixel point inside the stroke area and the pixel point outside the stroke area, the stroke polarity estimation module is A third calculation sub-module that calculates the stroke response strength by using the following formulas in two diagonal directions,

Here, w is one-eighth of the height of the original image, and f _n (i) is a third calculation submodule indicating the color value of pixel point i in channel n, and the calculated four stroke response intensities. The maximum stroke response intensity is the following two conditions, that is, the polarity of [f _n (i) −f _n (l)] and [f _n (i) −f _n (l)] in channel n A fourth determination submodule that determines whether or not two conditions are satisfied and that the stroke response strength is greater than a predetermined threshold value are satisfied, and the result of the fourth determination submodule is positive, a fourth specific sub-module that identifies the estimated stroke polarity of the text based on the polarity of [f _n (i) −f _n (l)] or [f _n (i) −f _n (k)]; and If the result of the fourth determination submodule is negative, the calculated stroke response strength is increased until a certain stroke response strength satisfies the two conditions. In order to trigger the fourth determination sub-module, or when the four stroke response intensities do not satisfy the two conditions, the pixel point i is set as a non-stroke pixel point. The apparatus of claim 11 comprising three trigger submodules.

ここで、ｗは原画像の高さの８分の一であり、f_n(i)はチャネルｎにおいて画素点iの色値を示す第４の算出サブモジュール、前記筆画応答強度が次の二つの条件、即ち、チャネルｎにおいて[f_n(i)−f_n(l)]と[f_n(i)−f_n(k)]との極性が一致し、且つ前記筆画応答強度が所定の閾値よりも大きいという二つの条件を同時に満たしているか否かを判断する第５の判断サブモジュール、前記第５の判断サブモジュールによる結果が肯定の場合に、[f_n(i)−f_n(l)]又は[f_n(i)−f_n(k)]の極性に基づいて前記原画素点iの初期極性を特定する第５の特定サブモジュール、四つの方向上の筆画応答強度の算出がすべて完了したか否かを判断する第６の判断サブモジュールと、前記第６の判断サブモジュールによる結果が肯定の場合に、四つの方向上の最大の筆画応答強度が対応する初期極性を、前記テキストの推定筆画極性として特定する第６の特定サブモジュール、及び、前記第６の判断サブモジュールによる結果が否定の場合に、前記の第４の算出サブモジュールをトリガする第４のトリガサブモジュールを含む、付記１１に記載の装置。 (Supplementary Note 15) When the polarity indicates the relationship of the magnitude of the color value between the pixel point inside the stroke area and the pixel point outside the stroke area, the stroke polarity estimation module is A fourth calculation submodule that calculates the stroke response strength of each original pixel point in the original image by adopting the following formula in any one of two diagonal directions,

Here, w is one-eighth of the height of the original image, f _n (i) is a fourth calculation submodule indicating the color value of pixel point i in channel n, and the stroke response intensity is In the channel n, [f _n (i) −f _n (l)] and [f _n (i) −f _n (k)] have the same polarity, and the stroke response intensity is predetermined. A fifth determination submodule that determines whether or not two conditions of greater than the threshold value are simultaneously satisfied. When the result of the fifth determination submodule is affirmative, [f _n (i) −f _n ( l)] or [f _n (i) −f _n (k)] Based on the polarity of the fifth specific submodule for specifying the initial polarity of the original pixel point i, the calculation of the stroke response intensity in four directions If the result of the sixth determination submodule and whether the sixth determination submodule is affirmative, the maximum stroke response in four directions is determined. A sixth specifying sub-module that specifies the initial polarity corresponding to the answer strength as the estimated stroke polarity of the text, and the fourth calculating sub-module when the result of the sixth determining sub-module is negative 12. The apparatus of claim 11 including a fourth trigger submodule that triggers.

  (Supplementary Note 16) The determination module may be configured to acquire a first magnitude relationship between the updated luminance value or / and the updated color value after filtering and the primary luminance value or / and the primary color value. The supplementary note 11 includes a module, and a seventh determination submodule that determines whether or not the first magnitude relationship and the second magnitude relationship indicated by the stroke polarity are matched. Equipment.

  (Supplementary note 17) A method in which a computer extracts text in an image, wherein the computer acquires an original image including at least one line of text, each neighborhood in a neighborhood set from any original pixel point in the original image Based on the direct difference degree and the indirect difference degree up to the pixel point, the original luminance value or / and the primary color value of each original pixel point is subjected to stroke two-dimensional filtering, and the updated luminance value after filtering of the original image or Obtaining the updated color value, replacing the corresponding original luminance value or / and the primary color value with the updated luminance value or / and the updated color value after filtering, respectively, and emphasizing the text corresponding to the original image Performing the steps of generating an image and extracting text in the text-enhanced image, wherein the range of the neighborhood set is The original pixel points as the center, and side length becomes square is w, w is less method than the height of the original image.

  (Supplementary Note 18) A device for extracting text in an image, a module for acquiring an original image including at least one line of text, and a direct difference from an arbitrary original pixel point in the original image to each neighboring pixel point in the neighborhood set The original luminance value or / and primary color value of each original pixel point is subjected to stroke two-dimensional filtering based on the degree and the indirect difference degree, and the updated luminance value or / and the updated color value after filtering of the original image are obtained. A module for obtaining a text-enhanced image corresponding to the original image by replacing the corresponding original luminance value or / and primary color value with the updated luminance value or / and updated color value after filtering, And a module for extracting text in the text enhanced image, wherein the range of the neighborhood set is centered on the original pixel point and is A device whose length is a square with w being smaller than the height of the original image.

Claims (10)

A method by which a computer emphasizes text in an image,
The computer is
A current image acquisition step of acquiring an original image including at least one line of text;
Based on the direct difference level and the indirect difference level from an arbitrary original pixel point in the original image to each neighboring pixel point in the neighborhood set, a stroke of the original luminance value or / and the primary color value of each original pixel point is determined. An update value acquisition step of performing dimension filtering to acquire an updated luminance value or / and an updated color value after filtering of the original image, and corresponding to the updated luminance value or / and the updated color value after filtering of the original image Replacing the original luminance value or the primary color value, and executing an enhanced image generation step of generating a text enhanced image corresponding to the original image,
The range of the neighborhood set is a square whose center is the original pixel point and whose side length is w, and w is smaller than the height of the original image. The update value acquisition step includes:
Performing the algebraic subtraction on the original luminance value or / and the primary color value between the original pixel point and each neighboring pixel point to obtain the direct difference degree;
Obtaining the indirect difference based on a gradient module from the original pixel point to each neighboring pixel point of the neighborhood set;
Calculating a luminance value and / or a color value weight value of each neighboring pixel point with respect to the original pixel point based on the direct difference degree and the indirect difference degree;
Adopting the following stroke two-dimensional filter equation, calculating the updated luminance value of the original pixel point,

Here, N (i) represents the neighborhood set of pixel point i, w (i, j) represents the weight value of the luminance value of the neighboring pixel point j with respect to the original pixel point i, and f (j) is the neighborhood A step that is a luminance value of pixel point j, and / or
Adopting the following stroke two-dimensional filter equation, calculating the updated color value of the original pixel point,

Here, w _n (i, j) represents the weight value of the color value of the neighboring pixel point j with respect to the original pixel point i in the n channel, and f _n (j) represents the color value of the neighboring pixel point j in the n channel. The method of claim 1, comprising the steps of: The computer further comprises:
A polarity estimation step for estimating a stroke polarity of text in the original image after the original image acquisition step, wherein the polarity is between a pixel point located inside the stroke area and a pixel point located outside the stroke area; A polarity estimation step showing the relationship between the primary luminance values or / and the primary color values, and
After the update value acquisition step, it is determined whether or not the updated luminance value or / and the updated color value after filtering are matched with the stroke polarity, and if the result is affirmative, the primary luminance value or / primary color value is determined. The method according to claim 1, wherein an update value determination step for replacing is executed. When the polarity indicates the relationship of the magnitude of the luminance value between the pixel point inside the stroke area and the pixel point outside the stroke area, the polarity estimation step includes:
In the width direction, the height direction, and the two diagonal directions of the original image, the following steps are used to calculate the stroke response strength, respectively.

Here, w is one-eighth of the height of the original image, f (i) is a step indicating the luminance value of the pixel point i, and the maximum stroke response strength among the calculated four stroke response strengths is The following two conditions, that is, [f (i) −f (l)] and [f (i) −f (k)] have the same polarity and the stroke response intensity is larger than a predetermined threshold value. It is determined whether or not two conditions are satisfied. If the result is affirmative, the text is estimated based on the polarities of [f (i) −f (l)] and [f (i) −f (k)]. When the stroke polarity is specified and negative, the four stroke response strengths are selected in order according to the magnitude until a certain stroke response strength satisfies the two conditions, and the determination step is executed. The method according to claim 3, further comprising the step of setting the pixel point i as a non-stroke pixel point when neither of the stroke response intensities satisfies the two conditions. When the polarity indicates the relationship of the magnitude of the luminance value between the pixel point inside the stroke area and the pixel point outside the stroke area, the polarity estimation step includes:
In one of the width direction, the height direction, and two diagonal directions of the original image, the step of calculating the stroke response intensity of each original pixel point in the original image by employing the following equation:

Here, w is one-eighth of the height of the original image, and f (i) is a step indicating the luminance value of the pixel point i.
The stroke response intensity is the following two conditions, that is, [f (i) −f (l)] and [f (i) −f (k)] have the same polarity, and the stroke response intensity is a predetermined threshold value. Is determined based on the polarity of [f (i) −f (l)] or [f (i) −f (k)]. Step of identifying the initial polarity of the original pixel point i and whether or not the calculation of the stroke response intensity in the four directions is all completed, and if the result is affirmative, the maximum stroke response in the four directions is determined. The method according to claim 3, further comprising: specifying an initial polarity corresponding to an intensity as an estimated stroke polarity of the text; and, if negative, repeating the stroke response intensity calculating step. When the polarity indicates a color value magnitude relationship between a pixel point inside the stroke area and a pixel point outside the stroke area, the polarity estimation step includes:
In the width direction, the height direction and the two diagonal directions of the original image, the following steps are used to calculate the stroke response strength of each original pixel point in the original image, respectively:

Here, w is one-eighth of the height of the original image, f _n (i) is a step indicating the color value of pixel point i in channel n, and the largest of the four stroke response intensities calculated. The stroke response intensity has the following two conditions, that is, in channel n, [f _n (i) −f _n (l)] and [f _n (i) −f _n (l)] have the same polarity, and It is determined whether or not the two conditions that the stroke response intensity is greater than a predetermined threshold are satisfied, and if affirmative, [f _n (i) −f _n (l)] or [f _n (i ) −f _n (k)], the estimated stroke polarity of the text is specified, and in the negative case, the four stroke response strengths are increased until a certain stroke response strength satisfies the two conditions. The selection step is executed in accordance with the determination and the determination step is executed, or when none of the four stroke response intensities satisfy the two conditions, the pixel point i is set as a non-stroke pixel point. Including step A method according to claim 3. When the polarity indicates a color value magnitude relationship between a pixel point inside the stroke area and a pixel point outside the stroke area, the polarity estimation step includes:
In one of the width direction, the height direction, and two diagonal directions of the original image, the step of calculating the stroke response intensity of each original pixel point in the original image by employing the following equation:

Where w is one-eighth of the height of the original image, and f _n (i) indicates the color value of pixel point i in channel n,
The stroke response intensity has the following two conditions, that is, in channel n, [f _n (i) −f _n (l)] and [f _n (i) −f _n (k)] have the same polarity, In addition, it is determined whether or not the two conditions that the stroke response intensity is greater than a predetermined threshold are satisfied at the same time, and if affirmative, [f _n (i) −f _n (l)] or [f _n (i) −f _n (k)] based on the polarity of the original pixel point i, and determining whether or not the calculation of the stroke response intensity in the four directions is completed, In the case of affirmation, the initial polarity corresponding to the maximum stroke response strength in four directions is specified as the estimated stroke polarity of the text, and in the case of negative, the step of calculating the stroke response strength is repeated. The method of claim 3. The update value determining step includes:
Obtaining a first magnitude relationship between the updated luminance value or / and the updated color value after filtering and the primary luminance value or / and the primary color value; and the first magnitude relationship; The method according to claim 3, further comprising the step of determining whether or not the second magnitude relationship indicated by the stroke polarity matches. A device for enhancing text in an image,
An acquisition module that acquires an original image containing at least one line of text,
Based on the direct difference level and the indirect difference level from an arbitrary original pixel point in the original image to each neighboring pixel point in the neighborhood set, a stroke of the original luminance value or / and the primary color value of each original pixel point is determined. A filter module that performs dimensional filtering and obtains an updated luminance value or / and an updated color value after filtering of the original image, and the corresponding original luminance value or A replacement module that replaces each of the primary color values and generates a text-enhanced image corresponding to the original image,
The range of the neighborhood set is a square centered on the original pixel point and the side length is w, where w is smaller than the height of the original image. A method for a computer to extract text in an image, comprising:
The computer is
Obtaining an original image containing at least one line of text;
Based on the direct difference level and the indirect difference level from an arbitrary original pixel point in the original image to each neighboring pixel point in the neighborhood set, a stroke of the original luminance value or / and the primary color value of each original pixel point is determined. Performing dimension filtering to obtain an updated luminance value or / and an updated color value after filtering of the original image;
Replacing the corresponding original luminance value or / and primary color value with the updated luminance value or / and updated color value after filtering, respectively, and generating a text enhanced image corresponding to the original image; and Perform steps to extract text in
The range of the neighborhood set is a square centered on the original pixel point and the side length is w, and w is smaller than the height of the original image.

Images