KR100692527B1 - Method for generating masking model for human visual system to embed watermark - Google Patents

Abstract

A method for creating a masking model adapted to an HVS(Human Visual System) in order to insert a watermark into a digital image is provided to insert the watermark or a finger print into the digital image most strongly within a range that the inserted watermark and finger print are not offensive to human sight. A color difference average value between a specific pixel in a mask and adjacent pixels of the specific pixel is calculated. Difference of color difference between the specific pixel and the adjacent pixels thereof is averaged and thereby a standard deviation value is calculated. Based on a value preliminarily set by filtering for detecting an edge, each pixel is distinguished into a flat area, a strong edge area and a texture area. A weight value for varying insertion strength is assigned to insertion of a watermark if color difference of the pixel belongs to a domain other than a domain of a preliminarily set value.

Description

How to create masking model for human visual system for watermark insertion {METHOD FOR GENERATING MASKING MODEL FOR HUMAN VISUAL SYSTEM TO EMBED WATERMARK}

1 is a conceptual diagram for calculating an average and a variance of a peripheral value of a watermark embedded pixel according to the present invention;

2 is a weighted function response curve corresponding to weighting according to the present invention;

FIG. 3A is a test image generated by the HVS masking model, and FIG. 3B is a figure showing the calculated watermark embedding strength of the test image.

The present invention relates to a method for embedding a watermark invisibly in a spatial region of a digital image. More particularly, the present invention relates to a human visual system (HVS) by applying an insertion weight differently to each region of an image. The present invention relates to a method for generating a masking model for a human visual system for embedding a watermark for generating a masking model.

As a method for copyright protection of multimedia contents on the Internet, there is a digital watermarking or fingerprinting technique that inserts / extracts copyright information. In general, a watermark is a signal like noise. Therefore, when the watermark is inserted, the quality deteriorates in the content because it is in a trade-off with the toughness of the watermark. Therefore, many image watermarking embedding techniques require a method of strongly inserting a watermark such as noise without being visually recognized.

However, according to the existing watermarking embedding method, there is a problem that the quality of the content is deteriorated due to a conflict between the content and the watermark, and there is a problem that the watermark is distorted when the watermarking is weakly inserted. .

Accordingly, the present invention has been made to solve the above-mentioned problems of the prior art, and an object of the present invention is to insert a watermark and a fingerprint into a spatial region of a digital image, while minimizing the quality degradation of the image after insertion, while inserting strength is high. To create a masking model for the human visual system for watermark insertion, which creates a masking model suitable for the human visual system (HVS) by applying different insertion weights to each area of the image to maximize the To provide.

In order to achieve the above object, a method of generating a masking model for a human visual system for embedding a watermark according to the present invention includes a color difference average value (Aver (i, j)) between a specific pixel in a mask and adjacent pixels of the specific pixel. And calculating a standard deviation (StD (i, j)), which is a value obtained by averaging the difference in color difference between the specific pixel and adjacent pixels, and filtering each edge based on a predetermined value through filtering for edge detection. Is divided into a flat area, a strong edge area, and a texture area, and weights for varying the insertion strength when the color difference is in an area other than the set value area for each pixel Characterized in that comprises the step of giving.

Human Visual System (HVS) is a technique for maintaining im-perceptibility while strongly inserting a watermark or fingerprint. If you only prioritize the quality of the content, you can use the method of inserting it very weakly throughout the image, but if you consider the robustness against distortion to some extent, you should insert it as strongly as possible. HVS includes a function that tells you where in the image you want to insert strongly.

The basic principle of operation of HVS is based on Weber's law. In other words, Weber's law is that when the stimulus is weak, the stimulus can be changed even if the next stimulus is slightly stronger. I can feel it. When Weber's law is applied to digital images, the small amount of change in an image that the human eye can perceive depends on the surrounding pixels, implying that even the same amount of change may not be noticeable depending on the surrounding value.

A typical HVS masking model can be expressed as a relationship to the marginal values, using the standard deviation from the marginal values, which are statistical calculations. That is, small and weakly inserted in a flat area having a small deviation from neighboring pixels in proportion to the deviation value, while being strongly inserted in an edge or texture area having a large deviation.

Hereinafter, a method of generating a masking model for a human visual system for embedding a watermark according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a conceptual diagram for calculating an average and a variance of peripheral values of a watermark embedded pixel according to the present invention.

In order to calculate the HVS, first, the average and standard deviation of color differences with surrounding values should be calculated. As shown in FIG. 1, as an example, a 3x3 mask is used to calculate the average and standard deviation ( It shows the process of calculating StD). The formula is as follows.

--- 식(1)

--- Equation (1)

--- 식(2)

--- Equation (2)

Aver (i, j) is the average value in the mask, and StD (i, j) is the average of the differences with the surrounding values. Here, the calculated StD (i, j) is used as a default value of the HVS masking model used in the present invention.

Secondly, the image should be divided into a flat area, a strong edge area, and a texture area.

--- 식(3)

--- Equation (3)

The conditional expression that divides each pixel of the image into three regions is shown in Equation (3).

The distinction condition of the strong edge is E (i, j) when the result of the prewitt filter commonly used for edge detection of an image is E (i, j). If the value is greater than or equal to a predetermined threshold, it is defined as a strong edge. The threshold is equal to Equation (3), where Aver (E) represents the mean of E (i, j) and StD (E) represents the standard deviation of E (i, j). As shown in Equation (3), the reason for classifying images is to apply an insertion strength differently to each pixel. Meanwhile, the prewitt filter detects an edge through a first derivative of an amount of brightness change, which is a gradient of pixel brightness, among various methods for detecting an edge.

Finally, in the generation of this HVS masking model, the weighted function is used as shown in Equation (4) to weight the insertion for the dark and very bright areas. In general, very dark or very bright areas are known to be less sensitive to human vision, so the weighting function WF (i) is

--- 식(4)

--- Equation (4)

The weighting function WF (i) defined above is determined by the average value Aver (i, j) of the pixels. In other words, if Aver (i, j) is a small value, a weight between 1 and 2 is applied because it is a dark area. On the contrary, even when Aver (i, j) is large, the weights between 1 and 2 are equally applied.

2 is a weight function response curve corresponding to a weight function application according to the present invention.

As shown in Fig. 2, the weight range used to weight the insertion in the dark and very bright areas is between 1 and 2, and the closer the Aver (i, j) is to 0 or 255, the closer to the 2 value. Output the weights.

Finally, the HVS used in the present invention is as shown in Equation (5) below.

--- 식(5)

--- Equation (5)

Here, K is a constant and is defined as 2. The K is the minimum value of the HVS, and may be defined differently according to the insertion strength or the type of the image.

FIG. 3A is a test image generated by the HVS masking model, and FIG. 3B is a figure showing the calculated watermark embedding strength of the test image.

As shown in FIG. 3B, the gray part is a part calculated by K, and the white part is a part which is strongly inserted according to StD (i, j).

As described above, the watermarking method according to the present invention applies the insertion strength of the watermark differently according to the area of the image, and the HVS masking model calculation method basically calculates the standard deviation value from the peripheral values of the pixels to which the watermark is to be inserted. The image is divided into three areas: flat area, texture area, and edge area, and the edge area is classified into strong edge and normal edge. In this way, the HVS masking is calculated to classify the image by region and to effectively insert a watermark in an invisible range.

Although the present invention has been described in more detail with reference to some embodiments, the present invention is not necessarily limited to these embodiments, and various modifications can be made without departing from the spirit of the present invention.

As described above, the method of generating a masking model for a human visual system for embedding a watermark according to the present invention, in inserting a watermark or a fingerprint into a digital image, is as powerful as possible in a range not unobtrusive to human vision. This has the effect of inserting a fingerprint. In addition, the present invention is easy to distinguish the image area by using the average of the difference with the peripheral value of the pixel as a basic change value in dividing the image into three areas (flat area, strong edge area, texture area).

Claims (5)

delete (a) Standard deviation StD which is an average value of the color difference average values Aver (i, j) between a specific pixel in an N × N mask and adjacent pixels of the specific pixel, and a color difference between the specific pixel and adjacent pixels. computing (i, j)) respectively; (b) distinguishing each pixel into a flat area, a strong edge area, and a texture area based on a predetermined value through filtering for edge detection; And (c) assigning a weight for each pixel to vary the insertion intensity when the color difference is in a region other than a set value region; Including, Masking model generation method for the human visual system for watermark embedding, characterized in that the division of each area in the step (b) is made by the following equation.

(Where N is a natural number of 2 or more, E (i, j) is the result of a prewitt filter, Aver (E) is the average of E (i, j), and StD (E) is E (i, j ), And M is an integer.) The method according to claim 2, (C) step weighting operation is a masking model generation method for a human visual system for watermark embedding, characterized in that the following formula.

(WF (i) is a weighted function.) delete The method according to claim 3, Masking model generation method for the human visual system for watermark embedding, characterized in that the result value of step (c) is defined by the following equation.

(Where K is a constant and HVS (i, j) represents a masking model function, respectively.)

Images