KR101875401B1 - A method for reversible data hiding based on pixel value prediction according to spatial locality and surface characteristics, a method for reversible watermarking using it, and an appratus for reversible data hiding, reversible water marking - Google Patents

Abstract

A data concealment method according to an embodiment of the present invention includes: receiving a cover image; Receiving confidential data to be hidden; Generating a predicted image based on spatial locality and surface characteristics from the cover image; Receiving a cover image sequence obtained from the cover image and a predicted image sequence obtained from the predicted image to generate a difference sequence; Generating a histogram corresponding to the difference sequence; Generating a shifted difference sequence by shifting the differential sequence according to shift information obtained from the high histogram, and embedding confidential data in a shifted difference sequence; Generating a hidden image sequence by combining the embedded sequence and the cover image sequence according to the embedding; And constructing a hidden image using the hidden image sequence and outputting the hidden image.

Description

FIELD OF THE INVENTION This invention relates to a reversible data concealment method, a reversible watermarking method, a reversible data concealment device, and a reversible watermarking device employing a pixel value prediction method based on spatial localization and surface characteristics. BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a reversible data concealment method, CHARACTERISTICS, A METHOD FOR REVERSIBLE WATERMARKING USING IT, AND AN APPRATUS FOR REVERSIBLE DATA HIDING, REVERSIBLE WATER MARKING}

The present invention relates to a data hiding method. More specifically, the present invention relates to a reversible data concealment method, a reversible watermarking method, a reversible data concealment apparatus, and a reversible watermarking apparatus applying a pixel value prediction method based on spatial localization and surface characteristics.

Recently, as various types of information are generated and circulated, important information is exposed to a third party due to hacking or transmission of information to the wrong path.

Especially, in case that important information such as military information or personal information is not applied to the information in the process of transmitting information, if such important information is transmitted to a third party, great damage may occur.

Recently, in order to prevent exposure of such important information, a data hiding technique for hiding important information on an image has been introduced.

Data hiding is a technique for embedding confidential data intended to be hidden in a cover medium such as an image or a moving picture and is used to store important digital information such as text, It is an indispensable technique for inserting the naked eye or the ear canal, and has recently taken an important place in applications such as digital library for medical, military and copyright protection.

Such a data concealment scheme should be reversible and there is a need to satisfy the imperceptibility requirement. That is, the original cover image and confidential data must be completely extracted from the hidden image generated by inserting confidential data into the cover image, and the confidential data embedded in the confidential image must not be recognized by the third party do.

Accordingly, various methods for satisfying the above requirements have been proposed, and in particular, the quality of the hidden image must be excellent in order to satisfy the non-recognition requirement.

However, in order to improve the quality of current hidden images, most of the proposed data hiding techniques cause distortion in the restored cover image obtained after extracting the confidential data from the hidden image due to the image processing on the cover image, Which is irreversible.

In order to solve this problem, a histogram of a difference sequence constituted by a difference between pixel values between adjacent pixels in a cover image is generated, and a differential sequence is shifted according to information of the histogram to generate a shifted difference sequence A secret sequence is generated by inserting confidential data into an element corresponding to a peak point of the histogram among the elements in the shifted difference sequence to generate an embedding sequence and composing the embedding sequence and the cover image sequence, A method of enabling reversible data concealment by generating a concealed image from a concealed image sequence has been proposed.

However, in the technique of inserting confidential data into a shifted difference sequence generated by shifting a differential sequence constituted by a difference of pixel values between adjacent pixels in the cover image, the number of bits of embeddable confidential data is shifted from the peak point of the histogram for the differential sequence There is a problem that there is a limit that is limited to the frequency. Accordingly, there is a problem that the amount of confidential data that can be concealed in the cover image is very limited by the frequency at the peak point of the histogram for the difference sequence.

SUMMARY OF THE INVENTION The present invention has been made in order to solve the above problems, and it is an object of the present invention to provide a method of generating a predictive image by precisely predicting pixel values using spatial locality and surface characteristics of a cover image, By constructing a difference sequence selectively applying the cover image sequence and the predictive image sequence according to whether the predictive pixel value is used or not for the predictive image sequence obtained from the predictive image and the cover image sequence obtained from the histogram of the difference sequence, The number of bits of inserting confidential data can be greatly increased, and thus, the amount of confidential data that can be hidden in the cover image can be greatly increased. Hiding method, reversible watermarking A reversible data concealment device, and a reversible watermarking device.

According to another aspect of the present invention, there is provided a data concealment method comprising: receiving a cover image; Receiving confidential data; Generating a predictive image from the cover page based on spatial locality and surface characteristics of the cover image; Receiving a cover image sequence obtained from the cover image and a predicted image sequence obtained from the predicted image to generate a difference sequence; Generating a histogram corresponding to the difference sequence; A confidential data inserting step of generating a shifted difference sequence by shifting the differential sequence according to the information of the histogram and generating an embedded sequence by embedding confidential data in a shifted difference sequence; Synthesizing the embedded sequence and the cover image sequence to generate a hidden image sequence; And outputting a concealed image from the concealed image sequence.

According to another aspect of the present invention, there is provided a data concealment apparatus comprising: a cover image input unit receiving a cover image; A confidential data input unit for inputting confidential data; A predicted image generating unit for generating a predicted image from the cover based on the spatial locality and surface characteristics of the cover image; A difference sequence generator for receiving a cover image sequence obtained from the cover image and a predicted image sequence obtained from the predicted image to generate a difference sequence; A histogram generator for generating a histogram corresponding to the difference sequence; A confidential data inserter for generating a shifted difference sequence by shifting the differential sequence according to the information of the histogram and generating an embedded sequence by embedding confidential data in a shifted difference sequence; A hidden image sequence generator for generating a hidden image sequence by combining the embedded sequence and the cover image sequence; And a hidden image output unit outputting a hidden image from the hidden image sequence.

According to another aspect of the present invention, there is provided a method of restoring confidential data, the method including receiving a hidden image, scanning a hidden image to generate a hidden image sequence, A prediction-based extraction mode determination step of determining an extraction mode in accordance with spatial locality and whether a surface property-based prediction value is used at a position of each element of the hidden image sequence; Restoring an embedding sequence corresponding to the hidden image sequence; Reconstructing a cover image sequence corresponding to the hidden image sequence according to the determined extraction mode; Extracting confidential data from the embedding sequence; And outputting a cover image from the cover image sequence.

According to another aspect of the present invention, there is provided an apparatus for restoring data, comprising: a hidden image receiver for receiving a hidden image and generating a hidden image sequence by scanning the hidden image; A prediction-based extraction mode determining unit for determining an extraction mode according to spatial locality and whether a surface property-based prediction value is used at a position of each element of the hidden image sequence; An embedding sequence restoring unit for restoring an embedding sequence corresponding to the hidden image sequence; A cover image sequence restoring unit for restoring a cover image sequence corresponding to the hidden image sequence according to the determined extraction mode; A confidential data extracting unit for extracting confidential data from the embedding sequence; And a cover image output unit for outputting a cover image from the cover image sequence.

In order to solve the above problems, a method according to an embodiment of the present invention may be embodied as a program stored in a recording medium and a recording medium in which the program is recorded, in order to execute the methods in a computer.

According to an embodiment of the present invention, after a predicted image is generated by predicting pixel values precisely using the spatial locality and surface characteristics of the cover image, a cover image sequence obtained from the cover image And a difference sequence in which a cover image sequence and a prediction image sequence are selectively applied depending on whether a prediction pixel value is used is generated for a prediction image obtained from a prediction image so that the frequency at a peak point of the histogram corresponding to the difference sequence is greatly increased So that the number of bits of insertable confidential data can be greatly increased so that the visual quality can be greatly increased to keep the amount of confidential data that can be hidden in the cover image, Apply pixel value prediction technique Reversible data hiding method, reversible watermarking method can provide a reversible data hiding apparatus and reversible watermarking unit.

1 is a conceptual diagram schematically showing an overall system according to an embodiment of the present invention.
2 is a block diagram illustrating a data hiding apparatus according to an embodiment of the present invention.
3 is a block diagram for explaining a data restoration apparatus according to an embodiment of the present invention.
4 is a flowchart for explaining a data hiding method according to an embodiment of the present invention.
FIG. 5 is a flowchart illustrating a method of generating a predictive image. FIG. 6 is a flowchart illustrating a method of generating a predictive image according to an exemplary embodiment of the present invention.
7 to 8 are illustrations showing a cover image, a predictive image generated from a cover image, and a hidden image embedded with confidential data.
FIG. 9 and FIG. 10 are flowcharts for explaining a data restoration method according to an embodiment of the present invention, and are illustrations for explaining a process thereof.

예를 들어, 위치 ①의 좌표는 (x_i, y_j)이고, 픽셀 값이 I, J, K, N인 픽셀의 좌표는 각각 (x_i+1, y_j-1), (x_i, y_j-1), (x_i-1, y_j-1), (x_i-1, y_j)이며, P(x, y)는 좌표 (x, y)에서의 픽셀 값을 나타낼 수 있다. I, J, K, N은 도 5 (A)에서 보는 바와 같이 155, 158, 158, 160의 값이 예시될 수 있다.
그리고 상기 수학식 1에서, 제1 임시 예측 픽셀 값을 구하는 위치와 인접해 있는 픽셀들이 제1 임시 예측 필셀 값에 미치는 영향을 나타내는 계수는 1로 하였고, 한 픽셀 떨어져 있는 위치의 픽셀들이 제1 임시 예측 픽셀 값에 미치는 영향을 나타내는 계수는 α로 나타낼 수 있다.
그리고, 예측 이미지 생성부(130)는 공간적 지역성의 정도를 판별하기 위하여 제1 임시 예측 픽셀 값과 주변 픽셀 값들과의 편차(dev)를 연산한다. 상기 편차 연산에 하기 수학식 2를 이용할 수 있다.

예를 들어, 예측 이미지 생성부(130)는 공간적 지역성 판별을 위한 기준 값을 임계 값 β로 설정할 수 있으며, 상기 제1 임시 예측 픽셀 값(V₁)과 그 주변 픽셀 값들간 편차(dev)를 이용하여 공간적 지역성의 정도를 판별할 수 있다.
이에 따라, 예측 이미지 생성부(130)는 상기 편차(dev)가 임계 값(β) 이상일 경우에는 해당 위치 주변의 영상은 픽셀 값 변화가 큰 경계 영역 혹은 공간적인 지역성이 낮은 영역으로 판단하여 픽셀 값 예측과정을 더 이상 수행하지 않고, 커버 이미지 내 해당 위치에서의 픽셀 값을 복사하여 예측 이미지를 생성할 수 있다.
반면, 예측 이미지 생성부(130)는 상기 편차(dev)가 임계 값(β) 미만이어서 공간적 지역성이 높은 경우, 해당 위치에서의 표면 특성을 추정하며, 상기 추정된 표면 특성에 따라, 제2 임시 예측 픽셀 값(V2)을 산출하고, 상기 표면 특성에 따라 상기 제1 임시 예측 픽셀 값 또는 상기 제2 임시 예측 픽셀 값을 선택적으로 적용하여 상기 예측 이미지를 생성할 수 있다.
그리고, 본 발명의 예측 이미지 생성부(130)는 역 s-순서(inverse s-order)로 상기 각 픽셀 별 예측 이미지 생성을 수행할 수 있다.
보다 구체적인 예측 이미지 생성 프로세스를 도 6을 참조하여 설명하면 다음과 같다.
먼저, 예측 이미지 생성부(130)는 예측 이미지 내 역 s-순서(inverse s-order)로 픽셀 위치를 선정하고(S301), 해당 위치가 픽셀 값 예측 가능 영역인지 여부를 판단한다(S303).
그리고, 예측 이미지 생성부(130)는 픽셀 값 예측 가능 영역이 아닌 경우, 픽셀 값 예측 과정을 수행하지 않고 커버 이미지 내 해당 위치에서의 픽셀 값을 복사하여 예측 이미지를 생성한다(S305).
한편, 픽셀 값 예측 가능 영역인 경우, 예측 이미지 생성부(130)는 상기 수학식 1을 이용하여 공간적 지역성에 기초한 제1 임시 예측 픽셀 값(V1)을 연산한다(S307).
그리고, 공간적 지역성의 정도를 판별하기 위하여 예측 이미지 생성부(130)는 수학식 2를 이용하여 제1 임시 예측 픽셀 값과 주변 픽셀 값들간의 편차(dev)를 연산한다(S309).
또한, 예측 이미지 생성부(130)는 공간적 지역성의 정도를 판별하기 위하여 상기 계산된 편차(dev)가 임계 값(β) 이상인지 여부를 판단한다(S311). 예측 이미지 생성부(130)는 상기 계산된 편차(dev)가 임계 값(β) 이상인 경우에는 공간적 지역성이 낮은 영역으로 판단하여, 픽셀 값 예측과정을 더 이상 수행하지 않고, 커버 이미지 내 해당 위치에서의 픽셀 값을 복사하여 예측 이미지를 생성한다(S305).
한편, 예측 이미지 생성부(130)는 상기 계산된 편차(dev)가 임계 값(β) 미만인 경우에는 해당 위치의 영상에서 픽셀 값 변화가 완만하여 공간적 지역성이 높은 것으로 판단하고, 커버 이미지 내 해당 위치에서의 표면 특성을 추정한다(S313). 추정된 표면 특성이 X축 방향으로의 단순 경사면(simple inclined surface)인지 또는 평면(flat surface)인지 여부를 판단하고(S315), 표면 특성이 X축 방향으로의 단순 경사면이거나 혹은 평면인 경우에는 해당 표면 특성을 잘 반영하는 제2 임시 예측 픽셀 값(V₂)을 예측 픽셀 값으로 결정하여 예측 이미지를 생성하고(S317), 다른 표면 특성인 경우에는 상기 제1 임시 예측 픽셀 값을 예측 픽셀 값으로 결정하여 예측 이미지를 생성한다(S319).
상기 흐름도에 따라 표면 특성을 추정하는 과정을 도 5에 도시된 도면을 참조하여 보다 구체적으로 설명하면, 전체적인 변수 설정 및 산출과정은 하기 수학식 3과 같다.

수학식 3과 같이, 임의의 두 지점 (x₁, y₁)과 (x₂, y₂)를 지나는 직선의 방정식은 y = {(y₂-y₁)/(x₂-x₁)}(x-x₁) + y₁ 이 될 수 있다. 이에 따라 상호 인접한 픽셀 사이의 직선방정식을 산출할 수 있다. 즉, 예측 이미지 생성부(130)는 픽셀 값이 K인 픽셀과 X축으로 인접한 픽셀 값이 J인 픽셀 사이에서, 픽셀 값에 대하여 X축 방향만을 고려한 직선방정식을 산출할 수 있다.
예측 이미지 생성부(130)에서는 (x_i-1, K)와 (x_i, J)를 지나는 직선의 방정식 P_L(x)를 수학식 3에서 개시된 식과 같이 산출할 수 있다. 동일한 방법으로 예측 이미지 생성부(130)는 픽셀 값이 J인 픽셀과 픽셀 값이 I인 픽셀 사이에서, 픽셀 값에 대하여 X축 방향만을 고려한 직선방정식 P_R(x)를 산출할 수 있다.
마찬가지로, 예측 이미지 생성부(130)는 픽셀 값이 N인 픽셀과 픽셀 값이 J인 픽셀 사이에서, 픽셀 값에 대하여 X축 방향만을 고려한 직선방정식 P_LD(x)를 수학식 3에 개시된 식과 같이 산출할 수 있다.
그리고, 예측 이미지 생성부(130)는 산출된 직선방정식들(P_L(x), P_R(x), P_LD(x))에 기초하여, 표면 특성 추정을 수행할 수 있다.
먼저, 예측 이미지 생성부(130)는 P_L(x), P_R(x), P_LD(x) 식들의 기울기가 모두 0보다 큰 경우에는 왼쪽보다 오른쪽의 픽셀 값이 큰 값을 갖는 단순 경사면(simple inclined surface)이라고 추정할 수 있다.
또한, 예측 이미지 생성부(130)는 P_L(x), P_R(x), P_LD(x) 식들의 기울기가 모두 0보다 적은 경우에는 오른쪽보다 왼쪽의 픽셀 값이 더 큰 값을 갖는 단순 경사면(simple inclined surface)이라고 추정할 수 있다.
그리고, 예측 이미지 생성부(130)는 P_L(x), P_R(x), P_LD(x) 식들의 기울기가 모두 0인 경우에는 왼쪽과 오른쪽이 평평한 평면(flat surface) 이라고 추정할 수 있다.
이에 따라, 예측 이미지 생성부(130)는 직선방정식의 기울기에 따라, 예측 위치에서 X축방향의 표면 특성이 단순 경사면(simple inclined surface) 혹은 평면(flat surface)으로 추정된 경우에는 해당 표면 특성을 잘 반영하는 제2 임시 예측 픽셀 값 V₂를 하기 수학식 4와 같이 연산하고, 상기 제2 임시 예측 픽셀 값 V₂를 위치 ①에서의 예측 픽셀 값으로 결정하여 예측 이미지를 생성 할 수 있다.

또한, 그 외의 표면 특성을 갖는 경우에는 예측 이미지 생성부(130)는 앞서 산출된 제1 임시 예측 값 V₁을 위치 ①에서의 예측 픽셀 값으로 결정하여 예측 이미지를 생성할 수 있다. 따라서 위치 ①에서의 예측 픽셀 값 V는 하기 수학식 5와 같이 결정될 수 있다.

결과적으로, 예측 이미지 생성부(130)는 예측 이미지 P를 상기 수학식 5에 의해 생성할 수 있다. 수학식 5에서, C(x_i, y_j)는 좌표 (x_i, y_j)에서 커버 이미지의 픽셀 값을 나타내며, V₁(x_i, y_j)과 V₂(x_i, y_j)는 좌표 (x_i, y_j)에서 제1 임시 예측 픽셀 값과 제2 임시 예측 픽셀 값을 각각 나타내고, P(x_i, y_j)는 좌표 (x_i, y_j)에서 예측 이미지의 픽셀 값을 나타낸다. 이에 따라, 도 5(C)의 위치 ④에서도 동일한 방법으로 예측 픽셀 값이 연산될 수 있어 예측 이미지가 생성될 수 있다.
이와 같은 방식에 의해, 예측 이미지 생성부(130)는 수학식 1 내지 수학식 5를 이용하여 예측 이미지를 생성할 수 있으며, 이에 따라 도 5 (B)에서 보는 바와 같이 예측 이미지의 상위 2개 행과 좌우측 각 2개 열은 픽셀 값 예측 가능 영역이 아니기 때문에 커버 이미지의 픽셀 값을 복사하여 사용할 수 있고, 나머지 영역이 픽셀 값 예측 가능 영역이 될 수 있다. 예측 이미지 생성부(130)는 도 5(B)의 픽셀 값 예측 가능 영역 내 역 s-순서에 따른 각 위치에서, 제1 임시 예측 픽셀 값을 수학식 1에 따라 계산하고, 수학식 2를 사용하여 계산된 편차(dev)를 임계 값(β)과 비교하여 공간적 지역성의 정도를 판별한다. 공간적 지역성이 낮을 경우, 픽셀 값 예측 과정을 더 이상 수행하지 않고 커버이미지 내 해당 위치에서의 픽셀 값을 복사하여 예측 이미지를 생성한다. 공간적 지역성이 높을 경우, 수학식 3을 적용하여 표면 특성에 따른 단순 경사면 또는 평면 여부를 결정하고, 표면 특성에 따라 제1 임시 예측 픽셀 값 혹은 제2 임시 예측 픽셀 값을 선택적으로 적용하여 예측 이미지를 생성하면, 도 5(D)의 예측 이미지가 생성 될 수 있다.
보다 구체적으로, 도 7 내지 도 8을 참조하면, 커버 이미지 입력부(110)는 커버 이미지를 입력 받고, 커버 이미지에 대하여 픽셀 값을 좌측에서 우측으로, 위에서 아래쪽으로 순차적으로 이루어지는 역 s-순서(inverse s-order)로 스캔(scan)하여 픽셀 값들로 구성되는 커버 이미지 시퀀스 C를 생성할 수 있다.
예측 이미지 생성부(130)는 생성된 예측 이미지를 역 s-순서(inverse s-order)로 스캔(scan)하여 예측 이미지 시퀀스 P를 생성할 수 있다.
그리고, 차분 시퀀스 생성부(140)는 상기 각 요소별 위치에서의 예측 픽셀 값 사용 여부 정보에 따라 커버이미지 시퀀스와 예측 이미지 시퀀스를 선택적으로 적용하여 차분 시퀀스 D를 생성할 수 있다.
차분 시퀀스 생성부(140)에서는 차분 시퀀스를 생성함에 있어서, 기존의 APD기법에 비하여 상기 예측 이미지 시퀀스와 예측 여부 정보를 추가적으로 이용함에 따라, 피크 포인트에서의 빈도수를 더욱 향상시키고, 이에 따라 더 많은 기밀 데이터가 삽입 가능한 차분 시퀀스가 생성될 수 있다.
차분 시퀀스 D의 각 요소를 D_i라고 하면, 차분 시퀀스 생성부(140)는 하기 수학식 6과 같이 커버 이미지 시퀀스의 요소와 예측 이미지 시퀀스의 요소를 이용하여 차분 시퀀스를 생성할 수 있다.

그리고, 데이터 은닉 장치(100)의 히스토그램 생성부(150)는 상기 차분 시퀀스에 대응하는 히스토그램을 생성한다. 여기서, 히스토그램 생성은 차분 시퀀스를 쉬프트 하는데 필요한 쉬프트 정보를 얻기 위한 것으로, 차분 시퀀스 쉬프트를 이용한 데이터 은닉기법(APD)을 적용하기 위한 것이며, 그 기술구성은 논문 Y. C. Li, C. M. Yeh, and C. C. Chang, Data hiding based on the similarity between neighboring pixels with reversibility, Digital Signal Processing, Vol. 20, No. 4, pp 1116-1128, July 2010 을 참조하여 서술될 수 있다.
히스토그램 생성부(150)는 차분 시퀀스에 대한 히스토그램을 생성하여 차분 시퀀스를 쉬프트(shift) 하기 위한 정보들 즉, 최대 빈도수를 갖는 피크 포인트 (PP₁, peak point 1), 두 번째로 큰 빈도수를 갖는 피크 포인트(PP₂, peak point 2), PP₁에 가장 가까이에 있는 빈도수가 0인 값을 갖는 CZP₁(closest zero point 1), PP₂에 가장 가까이에 위치한 빈도수가 0인 값을 갖는 CZP₂(closest zero point 2)를 결정할 수 있다.
이에 따라, 기밀 데이터 삽입부(160)는 히스토그램의 정보를 이용하여 차분 시퀀스를 하기 수학식 7과 같은 처리에 따라 쉬프트시켜 쉬프트된 차분시퀀스 DS를 생성할 수 있다.

그리고, 기밀 데이터 삽입부(160)는 하기 수학식 8과 같은 연산 처리에 따라 DS에 기밀 데이터를 삽입하여 임베딩 시퀀스 DE를 생성할 수 있다.

은닉 이미지 시퀀스 생성부(170)는, 상기 커버 이미지 시퀀스 C와 상기 임베딩 시퀀스 DE에 대하여 수학식 9와 같은 연산처리를 하여 최종적으로 은닉 이미지 시퀀스 S를 생성함으로써, 차분 시퀀스에 대한 히스토그램의 각 피크 포인트에 대응하는 픽셀들에 기밀 데이터를 비트별로 각각 삽입할 수 있게 된다.

이에 따라, 은닉 이미지 출력부(180)는 은닉 이미지 시퀀스를 스캔(scan)하여 역 s-순서(inverse s-order)로 은닉 이미지를 구성한 후, 이를 출력할 수 있다.
도 5, 도 7 내지 도 8은 커버 이미지, 커버 이미지로부터 생성된 예측 이미지, 기밀 데이터가 임베딩 된 은닉 이미지를 나타내는 예시도이다.
도 5, 도 7 내지 도 8에서는 이와 같은 전체 이미지 처리 과정에 의한 기밀 데이터 임베딩을 예시적으로 나타내고 있다.
먼저, 도 5, 도 7 내지 도 8에서는 상기 예측 이미지 생성시의 수학식 1과 수학식 2에서 사용된 계수 α를 0.298로 하고, 수학식 4에서 사용된 임계 값(β)을 15로 하여 생성된 예측 이미지의 예를 나타내고 있다. 이에 따라 도 5 (B)의 ①~⑧까지의 각 위치에서 제1 임시 예측 픽셀 값 V₁을 수학식 1에 따라 계산하면, 각각 157.0, 156.7, 157.3, 157.5, 158.0, 157.6, 158.7, 158.2 가 된다. 그리고 계산된 제1 임시 예측 픽셀 값 V1을 수학식 2에 적용하여 편차(dev)를 구하면, 각각 10.9, 10.3, 8.1, 7.3, 13.2, 8.4, 15.3, 12.2 가 된다.
이에 따라, 상기 임계 값(β)을 15로 설정하면, 도 5, 도 7 내지 도 8과 같이, 위치 ⑦에서의 편차(15.3)가 임계 값(β) 이상이기 때문에 예측 이미지 생성부(130)는 위치 ⑦을 공간적 지역성이 낮은 곳으로 판별하여 픽셀 값 예측과정을 더 이상 수행하지 않고, 커버 이미지 내 해당 위치에서의 픽셀 값(159)을 복사하여 예측 이미지를 생성할 수 있다.
또한, 예측 이미지 생성부(130)는 편차가 임계 값(β) 미만인 나머지 위치 ①~⑥, ⑧에서 영상의 표면 특성을 추정하여, 단순 경사면(simple inclined surface) 혹은 평면(flat surface)인 영역으로 추정되는 위치 ①, ④, ⑥을 판별할 수 있다.
그리고, 예측 이미지 생성부(130)는 위치 ①, ④, ⑥에서의 제2 임시 예측 픽셀 값 V₂를 수학식 4에 따라 계산하여, 각각 157.8, 158.0, 157.0을 산출하며, 이 값(제2 임시 예측 픽셀 값)을 반올림하여 각각의 위치에서의 예측 픽셀 값으로 결정하여 예측 이미지를 생성 할 수 있다.
한편, 예측 이미지 생성부(130)는 그 외의 위치 ②, ③, ⑤, ⑧에서의 표면 특성은 단순 경사면 혹은 평면이 아닌 것으로 추정되므로, 앞서 산출된 제1 임시 예측 픽셀 값 V₁을 반올림 한 후, 해당 위치에서의 예측 픽셀 값으로 결정하여 예측 이미지를 생성할 수 있다.
이러한 과정을 거쳐 구성된 예측 이미지가 도 5, 도 7 내지 도 8에서 설명될 수 있다. 도 8에서 나타나는 바와 같이, 본 발명의 실시 예에 따른 차분 시퀀스 D를 생성하고, 그 히스토그램을 생성하게 되면, PP₁=0, PP₂=1이며 h(PP₁)=12, h(PP₂)=3 인 히스토그램 정보가 생성될 수 있다. 이에 따라, 차분 시퀀스에 삽입되어 커버 이미지에 은닉 가능한 최대 데이터 비트수는 h(PP₁)과 h(PP₂)의 합인 15비트가 되며, 전술한 APD 방식에 비하여 4비트 이상 증가함을 확인할 수 있다. 또한, 임계 값(β)을 20로 설정하면, 위치 ⑦에서도 예측 픽셀 값이 사용되어 h(PP₁)=13, h(PP₂)=3이 되고, 삽입되는 기밀 데이터의 비트수가 1비트 더 증가 할 수 있다. 한편, 본 발명의 실시 예에서, 임계 값(β)을 0으로 설정하면 예측 이미지의 모든 픽셀은 커버 이미지의 픽셀 값을 그대로 복사한 값이 되므로, 임계 값(β)이 0인 경우에는 기존의 APD기법과 동일하게 동작할 수 있다.
도 9 내지 도 10은 본 발명의 실시 예에 따른 데이터 복원 방법을 설명하기 위한 흐름도와, 그 과정을 설명하기 위한 예시도 이다.

*도 9를 참조하면, 본 발명의 실시 예에 따른 데이터 복원 장치(200)는, 먼저 은닉 이미지 수신부(210)를 통해 은닉 이미지를 수신하고, 수신된 은닉 이미지를 좌측에서 우측으로, 위에서 아래쪽으로 역 s-순서(inverse s-order)로 스캔(scan)하여 픽셀 값들로 구성되는 은닉 이미지 시퀀스 S를 생성한다(S201).
그리고, 데이터 복원 장치(200)의 예측기반 추출 모드 결정부(220)는 상기 은닉 이미지 시퀀스의 각 위치 별로 공간적 지역성과 표면 특성을 추정하여 예측 픽셀 값 사용 여부를 판단하고, 예측 값 사용 여부에 따라 추출 모드를 결정한다(S203).
예측기반 추출 모드 결정부(220)는 은닉 이미지 시퀀스의 각 요소별 위치에서, 3가지 추출 모드 중에서 해당 추출 모드를 결정한다. 즉, 예측기반 추출 모드 결정부(220)는 영상의 상위 2개 행의 경우(i<imagewidth*2) 혹은 이전 위치와 현재 위치 모두에서 예측 픽셀 값이 사용되지 아니한 경우에는 추출 모드를 Mode 1으로 결정하고, 현재 위치에서 예측 픽셀 값이 사용된 경우에는 추출 모드를 Mode 2로 결정하며, 이전 위치에서 예측 픽셀 값이 사용되었으나 현재 위치에서는 예측 픽셀 값이 사용되지 아니한 경우에는 추출 모드를 Mode 3로 결정한다.
예측기반 추출 모드 결정부(220)의 동작을 자세히 기술하면 다음과 같다. 예측기반 추출 모드 결정부(220)는 도 10의 은닉 이미지 시퀀스 S의 각 요소별 위치에서 추출 모드를 결정할 수 있다.
현재 위치가 i인 경우, 임베딩 시퀀스는 DE_i까지 복원이 완료되었고, 커버 이미지 시퀀스는 S_i-1까지 복원이 완료되었다. 임베딩 시퀀스와 커버 이미지 시퀀스 복원에 대한 상세한 절차는 후술 하기로 한다.
먼저, 예측기반 추출 모드 결정부(220)는 위치 i가 픽셀 값 예측 가능 영역인지 여부를 판단한다. 픽셀 값 예측 가능 영역이 아닌 경우, 예측기반 추출 모드 결정부(220)는 해당 위치 i 에서 예측 픽셀 값이 사용되지 않은 것으로 판단할 수 있다.
현재 위치가 픽셀 값 예측 가능 영역인 경우, 예측기반 추출 모드 결정부(220)는 기 복원된 커버 이미지의 인접 픽셀 값들을 이용하여 수학식 1과 같이 공간적 지역성에 기초한 복원 제1 임시 예측 픽셀 값을 계산하고, 공간적 지역성의 정도를 판별하기 위하여, 복원 제1 임시 예측 픽셀 값과 인접 픽셀 값들과의 편차(dev)를 수학식 2와 동일한 방법으로 계산하여, 계산된 편차(dev)가 임계 값(β) 이상인지 여부를 판단한다.
예측기반 추출 모드 결정부(220)는 상기 계산된 편차(dev)가 임계 값(β) 이상이면 공간적 지역성이 낮은 경우이기 때문에 해당 위치 i 에서 예측 픽셀 값이 사용되지 아니한 것으로 판단 할 수 있다. 그리고 예측기반 추출 모드 결정부(220)는 상기 계산된 편차(dev)가 임계 값(β) 미만이면 공간적 지역성이 높은 경우이기 때문에 예측 픽셀 값이 사용된 것으로 판단할 수 있다. 예측기반 추출 모드 결정부(220)는 위치 i값을 증가 시키면서 이러한 과정을 반복적으로 수행하여, 이전 위치와 현재 위치에서 픽셀 값 사용 여부에 따라 추출 모드를 결정할 수 있다.
현재 위치에서 예측 픽셀 값이 사용되어 추출 모드가 Mode 2인 경우, 예측기반 추출 모드 결정부(220)는 해당 위치에서의 복원 예측 픽셀 값(RP_i)을 연산하기 위하여 은닉 방식에서와 동일한 방법으로 표면 특성을 추정한다. 예측기반 추출 모드 결정부(220)는 추정된 표면 특성이 단순 경사면(simple inclined surface)이거나 평면(flat surface)인 경우 기 복원된 커버 이미지 내 인접한 픽셀 값들을 사용하여 복원 제2 임시 예측 픽셀 값을 수학식 4와 동일한 방법으로 연산하여, 해당 위치에서의 복원 예측 픽셀 값(RP_i)으로 결정한다. 예측기반 추출 모드 결정부(220)는 추정된 표면 특성이 단순 경사면 혹은 평면이 아닐 경우에는 상기 복원 제1 임시 예측 픽셀 값을 해당 위치에서의 복원 예측 픽셀 값(RP_i)으로 결정할 수 있다.
또한, 예측기반 추출 모드 결정부(220)는 추출 모드가 Mode 2 혹은 Mode 3인 경우에는 커버 이미지 복원에 사용되는, i번째 위치에서 연관 값 AV_i (Associated Value at i^th position)를 하기 수학식 12를 사용하여 계산할 수 있다.
또한, 데이터 복원 장치(200)의 임베딩 시퀀스 복원부(230)와 커버 이미지 시퀀스 복원부(240)는 은닉 이미지 시퀀스로부터 임베딩 시퀀스를 복원하고, 각 위치에서 결정된 추출 모드에 따라 은닉 이미지 시퀀스로부터 커버 이미지 시퀀스를 복원한다(S205).
기밀 데이터 추출부(250)는 임베딩 시퀀스로부터 기밀 데이터를 추출하고 이를 출력한다(S207). 추출된 기밀 데이터는 손실 없이 완벽하게 복원되어, 은닉된 기밀 데이터와 동일하게 된다.
커버 이미지 출력부(260)는 커버 이미지 시퀀스를 스캔(scan)하여 역 s-순서(inverse s-order)로 커버 이미지를 구성한 후, 복원된 커버 이미지를 출력한다(S209).
예측기반 추출 모드 결정부(220), 임베딩 시퀀스 복원부(230), 커버 이미지 시퀀스 복원부(240), 기밀 데이터 추출부(250)의 보다 구체적인 동작 및 구현 방식은 다음과 같다.
은닉이미지 시퀀스 에서 위치 i값을 0에서부터 imagewidth*imageheight-1까지 증가시키며 예측기반 추출 모드 결정부(220), 임베딩 시퀀스 복원부(230) 및 커버 이미지 시퀀스 복원부(240)에서의 처리를 반복적으로 수행하면, 그림 10의 임베딩 시퀀스 DE와 커버 이미지 시퀀스 C를 복원할 수 있다.
위치 i에서 예측기반 추출 모드 결정부(220)가 결정한 추출 모드(Mode 1 ~ Mode 3)에 따라, 임베딩 시퀀스 복원부(230)와 커버 이미지 시퀀스 복원부(240)가 해당 위치 i에서, 다음과 같이 각 모드별 처리 과정을 수행하면, 임베딩 시퀀스 DE_i와 커버 이미지 시퀀스 C_i를 복원 할 수 있다.
Mode 1: 영상의 상위 2개 행의 경우(i≤imagewidth*2) 혹은 이전 위치와 현재 위치 모두에서 예측 픽셀 값이 사용되지 아니한 경우의 추출 모드. Mode 1에서는 전술한 차분 시퀀스를 쉬프트(shift)하여 기밀 데이터를 은닉하는 기법인 APD(Adjacent Pixel Difference) 방식과 동일하게 하기 수학식 10과 수학식 11을 사용하여 임베딩 시퀀스 DEi와 커버 이미지 시퀀스 C_i를 복원할 수 있다.

Mode 2: 현재 위치에서 예측 픽셀 값이 사용된 경우의 추출 모드. 상기 수학식 10을 사용하여 임베딩 시퀀스 DE_i를 복원하고, 하기 수학식 12를 사용하여 커버 이미지 시퀀스 C_i를 복원한다. RP_i는 복원과정 중, 위치 i에서의 추출 모드 판단 결과가 Mode 2인 경우, 위치 i에서 계산되는 복원 예측 픽셀 값으로, 은닉과정과 동일한 방법으로 계산될 수 있다. 또한, 연관 값 AV_i (Associated Value at i^th position)는 복원과정 중, 위치 i에서 추출 모드 판단 결과가 Mode 2 혹은 Mode 3인 경우, 수학식 12와 같이 계산되는 값으로 커버 이미지 복원에 사용되는 값이며, 후술할 Mode 3에서도 이용될 수 있다.

Mode 3: 이전 위치에서 예측 픽셀 값이 사용되었으나, 현재 위치에서는 예측 픽셀 값이 사용되지 아니한 경우의 추출 모드. 수학식 10을 사용하여 임베딩 시퀀스 DE_i를 복원하고, 하기 수학식 13을 사용하여 커버 이미지 시퀀스 C_i를 복원한다. RP_i-1은 복원과정 중, 위치 i-1에서의 추출 모드 판단 결과가 Mode 2인 경우, 위치 i-1에서 계산된 복원 예측 픽셀 값이다. AV_i는 상기 수학식 12에서 산출된 값이다.

보다 구체적으로, 기밀 데이터 추출부(250)는 임베딩 시퀀스 DE에 수학식 14를 적용하여 임베딩 시퀀스에 은닉되어 있는 기밀 데이터를 추출하여 출력 할 수 있다(S207). 추출된 기밀 데이터는 손실 없이 완벽하게 복원되기 때문에 은닉된 기밀 데이터와 동일하다.

그리고, 복원 장치(200)는 커버 이미지 출력부(260)를 통해 복원된 커버 이미지 시퀀스를 스캔하여 역 s-순서(inverse s-order)로 커버 이미지를 복원하고, 복원된 커버 이미지를 출력한다(S209).
상기와 같은 복원처리 과정이 도 10에 구체적으로 예시되어 있다. 도 10에 도시된 바와 같이, 은닉 이미지 시퀀스 S로부터, 임베딩 시퀀스 DE와 커버 이미지 시퀀스 C를 복원할 수 있으며, 임베딩 시퀀스로부터 은닉된 기밀 데이터를 추출하면, 원본 기밀 데이터인 "101011110110101"을 손실 없이 추출할 수 있게 된다. 추출된 기밀 데이터는 은닉된 기밀 데이터와 일치한다.
이와 같이, 차분 시퀀스 쉬프트를 이용하는 데이터 은닉 방식에 있어서, 본 발명의 실시 예에 따라 공간적인 지역성이 높은 영역의 표면 특성에 따라 정밀하게 픽셀 값 예측을 수행하고, 이에 따른 차분 시퀀스를 생성하면, 차분 시퀀스의 히스토그램에서 피크 포인트에서의 빈도수가 크게 증가하게 되어, 삽입 가능한 기밀 데이터의 비트수를 크게 증가시킬 수 있다. 또한, 임계 값(β) 조절에 따라서 은닉 이미지의 화질이 커버 이미지와 구별이 어려울 정도로 매우 우수하게 유지되면서도 다양한 레벨로 기밀 데이터를 은닉할 수 있는 연산방법을 제공할 수 있다. 또한, 은닉 이미지로부터 기밀 데이터와 원본 커버 이미지를 손실 없이 완벽하게 복원할 수 있어, 의료, 국방, 디지털 라이브러리 등의 응용분야에 매우 효과적으로 응용될 수 있다.
한편, 상술한 본 발명에 따른 방법들은 컴퓨터에서 실행되기 위한 프로그램으로 제작되어 컴퓨터가 읽을 수 있는 기록 매체에 저장될 수 있으며, 컴퓨터가 읽을 수 있는 기록 매체의 예로는 ROM, RAM, CD-ROM, 자기 테이프, 플로피디스크, 광 데이터 저장장치 등이 있다.
컴퓨터가 읽을 수 있는 기록 매체는 네트워크로 연결된 컴퓨터 시스템에 분산되어, 분산방식으로 컴퓨터가 읽을 수 있는 코드가 저장되고 실행될 수 있다. 그리고, 상기 방법을 구현하기 위한 기능 적인(functional) 프로그램, 코드 및 코드 세그먼트들은 본 발명이 속하는 기술분야의 프로그래머들에 의해 용이하게 추론될 수 있다.
또한, 이상에서는 본 발명의 바람직한 실시 예에 대하여 도시하고 설명하였지만, 본 발명은 상술한 특정의 실시 예에 한정되지 아니하며, 청구범위에서 청구하는 본 발명의 요지를 벗어남이 없이 당해 발명이 속하는 기술분야에서 통상의 지식을 가진 자에 의해 다양한 변형 실시가 가능한 것은 물론이고, 이러한 변형 실시들은 본 발명의 기술적 사상이나 전망으로부터 개별적으로 이해되어서는 안될 것이다.The following merely illustrates the principles of the invention. Therefore, the following description merely illustrates the principles of the present invention. Therefore, it is intended that the present invention cover the modifications and variations of this invention provided they come within the scope of the appended claims and their equivalents. Furthermore, all of the conditional terms and embodiments listed herein are, in principle, only intended for the purpose of enabling understanding of the concepts of the present invention, and are not to be construed as limited to such specifically recited embodiments and conditions do.
It is also to be understood that the detailed description, as well as the principles, aspects and embodiments of the invention, as well as specific embodiments thereof, are intended to cover structural and functional equivalents thereof. It is also to be understood that such equivalents include all elements contemplated to perform the same function irrespective of the currently known equivalents as well as the equivalents to be developed in the future, i.e., the structure.
Thus, for example, it should be understood that the block diagrams herein illustrate conceptual aspects of exemplary circuits embodying the principles of the invention. Similarly, all flowcharts, state transition diagrams, pseudo code, and the like are representative of various processes that may be substantially represented on a computer-readable medium and executed by a computer or processor, whether or not the computer or processor is explicitly shown .
The functions of the various elements shown in the drawings, including the functional blocks shown in a processor or similar concept, may be provided by use of dedicated hardware as well as hardware capable of executing software in connection with appropriate software. When provided by a processor, the functions may be provided by a single dedicated processor, a single shared processor, or a plurality of individual processors, some of which may be shared.
Also, the explicit use of terms such as processor, control or similar concepts should not be construed exclusively to cite hardware capable of executing software, but rather may be embodied in the form of digital signal processor (DSP) hardware, ROM ), RAM (RAM), and non-volatile memory. Hardware according to known techniques may also be included.
In the claims hereof, the elements represented as means for performing the functions described in the detailed description are intended to include, for example, a combination of circuit elements or firmware / microcode etc. performing the function, Lt; RTI ID = 0.0 &gt; circuitry < / RTI &gt; It is to be understood that the invention defined by the appended claims is not to be construed as encompassing any means capable of providing such functionality, as the functions provided by the various listed means are combined and combined with the manner in which the claims require .
BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description of the present invention when taken in conjunction with the accompanying drawings, in which: There will be. In the following description, well-known functions or constructions are not described in detail since they would obscure the invention in unnecessary detail.
Hereinafter, a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.
1 is a conceptual diagram schematically showing an overall system according to an embodiment of the present invention.
Referring to FIG. 1, a system according to an embodiment of the present invention may include a data hiding apparatus 100 and a data restoring apparatus 200.
The data concealment apparatus 100 described below receives a cover image (or cover medium) such as an original image or an original moving image, receives confidential data to be concealed, and applies a pixel value prediction technique to convert a large amount of confidential data into reversible A server device, a terminal device, a smart phone, a PDA, a tablet, a printer, and the like, which are inserted into the cover image to generate and output a hidden image.
In addition, the data restoration apparatus 200 receives the hidden image in which a large amount of confidential data is inserted, and completely restores the large amount of confidential data and the cover image by applying a pixel value prediction method from the hidden image, A computer, a server device, a terminal device, a smart phone, a PDA, a tablet, a printer, and the like.
The confidential data may include various types of data that the user desires to conceal. Herein, the confidential data may be composed of data in a bit stream format, and may be inserted into a cover image by a reversible data concealment method using a pixel value prediction method in a form that is not recognized by a human. Accordingly, confidential information such as military information, medical information, and digital library information can be concealed on the cover image. When confidential data is concealed on the cover image, the concealed image may be referred to as a stego image, and the appearance may be slightly different from that of the cover image.
Also, according to the embodiment of the present invention, the confidential data may include watermark data including owner information or original confirmation information of the cover image. In this case, it is obvious that the data hiding apparatus 100 and the restoring apparatus 200 may be referred to as a watermarking apparatus 100 or a watermark restoring apparatus 200, respectively, And the like. Further, when including watermark data, the blind image may be referred to as a watermark image.

In such a system configuration, the data concealment apparatus 100 generates a predicted image based on the spatial locality and surface characteristics from the cover image, and generates a predicted image sequence obtained from the predicted image, Generating a difference sequence by using a cover image sequence obtained from the cover image, shifting the difference sequence according to information of a histogram corresponding to the difference sequence to generate a shifted difference sequence, Generates a hidden image sequence by compositing the embedded image sequence with the embedded image sequence, and outputs the hidden image using the hidden image sequence.
The data restoration apparatus 200 receives the hidden image, scans the hidden image in an inverse s-order to generate a hidden image sequence, and stores the position of each element of the hidden image sequence Determines an extraction mode according to whether the spatial locality and the surface property based prediction value are used, restores the embedding sequence from the hidden image sequence, restores the cover image sequence from the hidden image sequence according to the determined extraction mode, Extracts confidential data from the restored cover image sequence, and outputs the restored cover image from the restored cover image sequence.
According to such a configuration, the data concealment and restoration system using the concealment apparatus 100 and the restoration apparatus 200 according to the embodiment of the present invention can generate a concealed image difficult to visually distinguish from the cover image, The cover image and the confidential data hidden can be perfectly extracted without loss, and the concealed data amount can be greatly improved compared with the conventional method.
This will be described in more detail with reference to the following drawings.
2 is a block diagram illustrating a data hiding apparatus according to an embodiment of the present invention.
2, the data concealment apparatus 100 according to the embodiment of the present invention includes a cover image input unit 110, a confidential data input unit 120, a predicted image generation unit 130, a difference sequence generation unit 140, A histogram generator 150, a secret data inserter 160, a hidden image sequence generator 170, and a hidden image output unit 180.
First, the cover image input unit 110 receives a cover image for inserting confidential data from the outside, and scans the cover image in an inverse s-order to generate a cover image sequence. The confidential data input unit 120 receives confidential data to be hidden from the outside. The cover image and the confidential data can be received from various input media and communication devices and can be converted into a form that can be processed by the data concealment apparatus 100.
The predictive image generating unit 130 generates a predictive image based on spatial locality and surface characteristics from the input cover image and outputs the predictive image in inverse s-order, To generate a predicted image sequence.
More specifically, the predictive image generator 130 performs the following process to obtain a predictive pixel value at each position on the inverse s-order in the predictive image. A first temporary predictive pixel value is generated from pixel values around the position in the cover image based on the spatial locality and a degree of spatial locality is determined from the deviation between the first temporary predictive pixel value and the neighboring pixel values, Determining a second temporary predicted pixel value according to the surface characteristic, and calculating the second temporary predicted pixel value based on the spatial locality and the surface characteristic, The predictive image can be generated by selectively applying the second temporary predictive pixel value and the pixel value of the cover image.
The difference sequence generator 140 generates a difference sequence using the cover image sequence and the prediction image sequence.
The difference sequence generator 140 may generate a difference sequence by selectively applying the cover image sequence and the predictive image sequence according to whether prediction values are used or not. For example, when the predicted pixel value is not used according to the cover image sequence and the predictive pixel value use / non-use information of the predictive image sequence at each element position, the pixel value difference between adjacent pixels of the cover image And generates a difference sequence including a difference value between the cover image sequence and the predicted image sequence when the predicted pixel value is used.
Then, the histogram generator 150 generates a histogram corresponding to the difference sequence. Here, the histogram is for analyzing the peak point (PP) and the closest zero point (CZP), which is information necessary to shift the difference sequence.
Accordingly, the confidential data inserting unit 160 generates the differential sequence shifted by shifting the differential sequence based on the shift information obtained from the histogram, and sequentially outputs the confidential data to each peak point of the shifted difference sequence To generate an embedding sequence.
The hidden image sequence generator 170 combines the embedded sequence and the cover image sequence to generate a hidden image sequence.
Then, the hidden image output unit 180 scans the hidden image sequence to construct a hidden image in an inverse s-order, and then converts and outputs the hidden image into an appropriate form according to the output medium .
For example, the hidden image output unit 180 may include a display device for displaying image information, a transmission device for transmitting a hidden image, a management device for storing and managing a hidden image, or a printing device for outputting a sheet on which a hidden image is printed , And a variety of output devices for outputting a hidden image.
The more specific operations and implementations of the prediction image generator 130, the difference sequence generator 140, the histogram generator 150, the confidential data inserter 160, and the hidden image sequence generator 170, 4 to 8 will be described later.
3 is a block diagram for explaining a data restoration apparatus according to an embodiment of the present invention.
3, the restoration apparatus 200 according to the embodiment of the present invention includes a hidden image receiving unit 210, a prediction based extraction mode determining unit 220, an embedded sequence restoring unit 230, a cover image sequence restoring unit 230, A confidential data extracting unit 250, and a cover image outputting unit 260.
The hidden image receiving unit 210 receives the hidden image to be restored from the outside, and scans the hidden image in an inverse s-order to generate a hidden image sequence. The concealed image can be received from various input media or communication devices and can be converted into a form that can be processed by the data decompression apparatus 200.
Then, the prediction-based extraction mode determination unit 220 determines an extraction mode according to the spatial locality and the use of the surface property-based prediction value at the position of each element of the hidden image sequence.
More specifically, the prediction-based extraction mode determination unit 220 can identify whether pixel values can be predicted at each element position on the inverse s-order of the hidden image sequence, And whether or not the predicted value is used according to the spatial locality and the surface property discrimination, and determine an extraction mode at each element position of the hidden image sequence based on the discrimination. The procedure for identifying whether to use the predictive value at the position of each element will be described later with reference to FIG. 9 to FIG.
For example, when the prediction-based extraction mode determination unit 220 determines that the image is in the upper two rows of the hidden image (i If the prediction pixel value is not used at the previous position and the current position of the hidden image sequence, the extraction mode can be determined as Mode 1.
The prediction-based extraction mode determination unit 220 may determine the extraction mode to be Mode 2 when the prediction pixel value is used at the current position of the hidden image sequence.
In addition, the prediction-based extraction mode determination unit 220 may determine the extraction mode to be Mode 3 when the prediction pixel value is used at the previous position of the hidden image sequence but the prediction pixel value is not used at the current position.
The embedding sequence restoring unit 230 restores the embedding sequence synthesized in the hidden image.
For example, the embedding sequence reconstructing unit 230 may perform adaptive equalization (ADD) on the APD (Adjacent Pixel Difference, YC Li, CM Yeh, and CC Chang, Data Hiding Based on Similarity between Neighboring Pixels with Reversibility, Digital Signal Processing, Vol. 20, No. 4, pp. 1116-1128, July 2010) technique can be used to restore the embedding sequence.
The cover image sequence restoring unit 240 restores the original cover image from the hidden image.
The cover image sequence restoring unit 240 can restore the cover image sequence in the same manner as the APD method in Mode 1. In Mode 2 and Mode 3, the restored predicted pixel value and the associated value Can be used to restore the cover image sequence.
Also, the confidential data extracting unit 250 can easily extract confidential data from the embedding sequence. The confidential data extracted here is the same as the confidential confidential data because it is completely restored without loss.
More specific operations and implementations of the prediction-based extraction mode determining unit 220, the embedded sequence restoring unit 230, the cover image sequence restoring unit 240, and the classified data extracting unit 250 are shown in FIGS. 9 to 10 Again, this will be described later.
The cover image output unit 260 sequentially scans the pixel values of the cover image sequence, composes the cover image in inverse s-order, and outputs the cover image. Here, the restored cover image is the same as the original cover image because it is completely restored without loss.
Here, the cover image output unit 260 may include a display for displaying image information, a transmitting apparatus for transmitting a cover image, a management apparatus for storing and managing a cover image, or a printing apparatus for outputting a sheet on which a cover image is printed, And may be implemented as an output device of various types for outputting a cover image.
FIGS. 4 to 8 are flowcharts for explaining a data hiding method according to an embodiment of the present invention, and an example for explaining the process.
Hereinafter, the data concealment process will be described in more detail with reference to FIG. 4 to FIG.
4, a data hiding apparatus 100 according to an exemplary embodiment of the present invention includes a cover image input unit 110, The cover image is received, and the cover image is scanned in an inverse s-order to generate a cover image sequence (S101).
Thereafter, confidential data to be concealed is input to the cover image through the confidential data input unit 120 (S103).
The predictive image generator 130 determines whether the current position on the inverse s-order in the predictive image is a pixel value predictable area (S105).
If the current position is not a pixel value predictable area, the prediction image generation unit 130 generates a prediction image by copying the pixel value at the corresponding position in the cover image without performing the pixel value prediction. If the current position is a pixel value predictable area, the predictive image generator 130 may generate the predictive image in inverse s-order based on the spatial locality and the surface characteristic from the pixel values in the cover image have.
More specifically, when the current position is a pixel value predictable area, the predictive image generator 130 calculates a first temporary predictive pixel value from pixel values of a predetermined area in the cover image based on the spatial locality (S107). In order to determine the degree of spatial locality, the predicted image generation unit 130 calculates a deviation between the first temporary predicted pixel value and its surrounding pixel values (S109) (S111).
If the deviation is greater than or equal to the threshold value, the predictive image generation unit 130 determines that the spatial locality is low, and does not perform the pixel value prediction process any more, and generates a predicted image by copying the pixel value at the corresponding position in the cover image . If the deviation is less than the threshold value, the prediction image generating unit 130 determines that the spatial locality is high and estimates the surface characteristic at the corresponding position (S113).
When the surface characteristic is a simple slope or a simple flat surface, the predictive image generation unit 130 generates a second temporary predictive pixel value that reflects the surface characteristic and determines the predicted pixel value as a predictive pixel value to generate a predictive image . If the surface characteristic is not a simple slope or a plane, the predictive image generator 130 determines the first temporary predictive pixel value as a predictive pixel value to generate a predictive image (S115).
Then, the predictive image generator 130 scans the generated predictive image in an inverse s-order to generate a predictive image sequence (S117).
Thereafter, the difference sequence generator 140 selectively generates a difference sequence by selectively applying the cover image sequence and the predicted image sequence according to the predictive value use / non-use information at each element position (S119).
Then, the histogram generator 150 generates a histogram corresponding to the difference sequence (S121).
The confidential data inserting unit 160 generates a shifted difference sequence by shifting the differential sequence according to the shift information obtained from the histogram, and embeds the confidential data in the shifted difference sequence to generate an embedded sequence (S123).
The hidden image sequence generator 170 generates a hidden image sequence by combining the cover image sequence and the embedded sequence (S125).
Then, the hidden image output unit 180 scans the hidden image sequence to construct a hidden image in the reverse s-order, and outputs the hidden image (S127).
In general, spatial images can have spatial locality. Also, surface characteristics may exist in each part of the image. Accordingly, adjacent pixel values may have characteristics having very similar values, and pixels in a certain region may have various surface characteristics such as a curved surface and a flat surface.
Accordingly, the data concealment apparatus 100 according to the embodiment of the present invention selects a position that is expected to have a high localization property by using adjacent pixel values, and precisely sets the pixel value according to the surface property at the highly localized position And a predictive image sequence obtained from the predictive image is selectively applied according to the prediction pixel value use information at the position of each element By constructing the difference sequence by generating the difference value, the frequency at the peak point of the histogram can be greatly increased, and the number of bits of the confidential data inserted into the cover image can be greatly increased. As the pixel value prediction is more accurate, the frequency at the peak point of the histogram for the difference sequence constituted by the difference of the pixel values is greatly increased, so that the number of bits of confidential data inserted into the cover image can be greatly increased.
FIG. 5 is a flowchart illustrating a method of generating a predictive image. FIG. 6 is a flowchart illustrating a method of generating a predictive image according to an exemplary embodiment of the present invention.
5 shows a cover image, a pixel value prediction order, twelve surrounding pixels used for prediction, and a prediction image generated according to a prediction of the prediction image generating unit 130.
The predicted image generating unit 130 generates a predicted image using the pixel values of the cover image shown in FIG. 5 (A), using the predicted area (for example, the upper two rows of the predicted image and the pixels excluding two pixels of each of the left and right sides) The inverse s-order shown in Fig. 6 (B) can be used to predict the pixel values inside the bold line corresponding to the inverse s-order.
5 (C), the predicted image generating unit 130 calculates the predicted pixel value at each position of the predictable area using surrounding pixel values to generate a predicted image as shown in FIG. 5 (D) . In FIG. 5D, the underlined pixel values indicate the predicted pixel values at the corresponding positions, and the value 159 at the position 7 is not a predicted value but a pixel value at the corresponding position in the cover image.
Also, in the case of the pixel value predictable area, the prediction image generator 130 may use the 12 pixel values scanned in inverse s-order in pixel value prediction, 1 temporary prediction pixel value. For example, in Fig. 5 (C), the position (x _i , y _j ) Can be obtained by the following equation (1). &Lt; EMI ID = 1.0 &gt; Here, pixel values scanned in inverse s-order are used to completely restore the confidential data and the original cover image in the restoration process.

For example, the coordinates of position 1 are (x _i , y _j ), And the coordinates of the pixel whose pixel value is I, J, K, N are (x _{i + 1} , y _j-1 ), (x _i , y _j-1 ), (x _i-1 , y _j-1 ), (x _i-1 , y _j ), And P (x, y) can represent a pixel value at coordinates (x, y). I, J, K, and N may be values of 155, 158, 158, and 160 as shown in FIG. 5 (A).
In Equation (1), a coefficient indicating the influence of the pixels adjacent to the position for obtaining the first temporary predictive pixel value on the first temporary predictive pixel value is 1, The coefficient indicating the influence on the predicted pixel value can be represented by?.
The predictive image generator 130 calculates a deviation (dev) between the first temporary predictive pixel value and neighboring pixel values to determine the degree of spatial locality. The following equation (2) can be used for the deviation calculation.

For example, the prediction image generator 130 may set a reference value for determining the spatial locality to a threshold value beta, and the first temporary predicted pixel value V _One ) And the deviation (dev) between neighboring pixel values can be used to determine the degree of spatial locality.
Accordingly, when the deviation dev is equal to or greater than the threshold value beta, the prediction image generation unit 130 determines that the image around the position is a boundary region having a large pixel value change or a region having low spatial locality, The predicted image can be generated by copying the pixel value at the corresponding position in the cover image without further performing the prediction process.
On the other hand, if the deviation dev is less than the threshold value beta, the predicted image generation unit 130 estimates the surface characteristic at the corresponding position when the spatial locality is high, and, based on the estimated surface characteristic, The predictive pixel value V2 may be calculated and the predictive image may be generated by selectively applying the first temporary predictive pixel value or the second temporary predictive pixel value according to the surface characteristic.
The predictive image generator 130 of the present invention can perform predictive image generation for each pixel in an inverse s-order.
A more detailed prediction image generation process will be described with reference to FIG.
First, the predictive image generator 130 selects a pixel position in an inverse s-order (S301) and determines whether the corresponding position is a pixel value predictable area (S303).
If the predicted image is not a pixel value predictable area, the predicted image generating unit 130 generates a predicted image by copying the pixel value at the corresponding position in the cover image without performing the pixel value predicting process (S305).
Meanwhile, in the case of the pixel value predictable area, the predictive image generator 130 calculates the first temporary predictive pixel value V1 based on the spatial locality using Equation 1 (S307).
In order to determine the degree of spatial locality, the predictive image generator 130 calculates a deviation (dev) between the first temporary predictive pixel value and neighboring pixel values using Equation (2) (S309).
In addition, the predictive image generator 130 determines whether the calculated deviation dev is equal to or greater than a threshold value (beta) in order to determine the degree of spatial locality (S311). When the calculated deviation dev is equal to or greater than the threshold value beta, the prediction image generation unit 130 determines that the spatial locality is low, and does not perform the pixel value prediction process any more, And generates a prediction image (S305).
On the other hand, when the calculated deviation dev is less than the threshold value?, The predicted image generation unit 130 determines that the pixel value change is gradual and the spatial locality is high in the image of the corresponding position, (S313). &Lt; / RTI &gt; It is determined whether the estimated surface characteristic is a simple inclined surface or a flat surface in the X-axis direction (S315). If the surface characteristic is a simple inclined surface in the X-axis direction or a flat surface, The second provisional predicted pixel value (V ₂ ) Is determined as a predicted pixel value to generate a predicted image (S317). In case of another surface characteristic, the first temporary predicted pixel value is determined as a predicted pixel value to generate a predicted image (S319).
The process of estimating the surface characteristics according to the flowchart will be described in more detail with reference to the drawing of FIG.

As shown in Equation (3), at any two points (x _One , y _One ) And (x ₂ , y ₂ ) Can be expressed as y = {(y ₂ -y _One ) / (x ₂ -x _One )} (xx _One ) + y _One . Accordingly, a linear equation between adjacent pixels can be calculated. That is, the predictive image generation unit 130 can calculate a linear equation that considers only the X-axis direction with respect to pixel values between a pixel having a pixel value of K and a pixel having a pixel value of J adjacent to the X-axis.
The prediction image generating unit 130 generates (x _i-1 , K) and (x _i , J), the straight line equation P _L (x) can be calculated as shown in Equation (3). In the same way, the predictive image generation unit 130 generates a predicted image P (i, j) between a pixel having a pixel value J and a pixel having a pixel value I, _R (x) can be calculated.
Likewise, the predictive image generation unit 130 generates a predictive image P between a pixel having a pixel value N and a pixel having a pixel value J by using a linear equation P _LD (x) can be calculated as shown in Equation (3).
The predictive image generator 130 then outputs the calculated linear equations P _L (x), P _R (x), P _LD (x)). &lt; / RTI &gt;
First, the predicted image generation unit 130 generates P _L (x), P _R (x), P _LD (x) are all greater than 0, it can be assumed that the pixel value on the right side of the left side is a simple inclined surface having a large value.
In addition, the prediction image generating unit 130 generates prediction image P _L (x), P _R (x), P _LD (x) are all less than 0, it can be assumed that the pixel value on the left side of the right side is a simple inclined surface having a larger value.
Then, the predicted image generating unit 130 generates P _L (x), P _R (x), P _LD (x) are all 0, it can be assumed that the left and right sides are flat surfaces.
Accordingly, when the surface characteristic in the X-axis direction at the predicted position is estimated as a simple inclined surface or a flat surface according to the slope of the linear equation, The second provisional predicted pixel value V ₂ Is calculated according to the following equation (4), and the second temporary predicted pixel value V ₂ Can be determined as a predicted pixel value at position 1, and a predicted image can be generated.

In addition, in the case of having other surface characteristics, the predictive image generation unit 130 generates the predicted first temporary predicted value V _One Can be determined as the predicted pixel value at position 1, and a predicted image can be generated. Therefore, the predicted pixel value V at position 1 can be determined as shown in Equation (5).

As a result, the predicted image generating unit 130 may generate the predicted image P according to Equation (5). In Equation 5, C (x _i , y _j ) Is the coordinate (x _i , y _j ) Represents the pixel value of the cover image, and V _One (x _i , y _j ) And V ₂ (x _i , y _j ) Is the coordinate (x _i , y _j ) Represents a first provisional predicted pixel value and a second provisional predicted pixel value, respectively, and P (x _i , y _j ) Is the coordinate (x _i , y _j ) Represents the pixel value of the predicted image. Accordingly, the predicted pixel value can be calculated in the same way at the position 4 in Fig. 5 (C), and a predicted image can be generated.
In this way, the predictive image generator 130 can generate a predictive image using Equations 1 to 5, and accordingly, as shown in FIG. 5B, And the left and right two columns are not pixel value predictable areas, the pixel value of the cover image can be copied and used, and the remaining area can be a pixel value predictable area. The predictive image generation unit 130 calculates the first temporary predicted pixel value at each position in the s-order in the pixel value predictable area of FIG. 5B according to Equation (1), and uses Equation 2 And compares the calculated deviation (dev) with the threshold value (beta) to determine the degree of spatial locality. If the spatial locality is low, the predicted image is generated by copying the pixel value at the corresponding position in the cover image without further performing the pixel value predicting process. If the spatial locality is high, a simple slope or a flat surface depending on the surface characteristics is determined by applying Equation (3), and a first temporary predictive pixel value or a second temporary predictive pixel value is selectively applied according to the surface characteristics, Once created, the predicted image of Figure 5 (D) can be generated.
7 to 8, the cover image input unit 110 receives a cover image, and outputs a pixel value to the cover image in inverse-s-order (left-to-right) s-order to generate a cover image sequence C consisting of pixel values.
The predictive image generator 130 may generate a predictive image sequence P by scanning the generated predictive image in an inverse s-order.
The difference sequence generator 140 may generate the difference sequence D by selectively applying the cover image sequence and the predicted image sequence according to the predictive pixel value use information at each element position.
The difference sequence generation unit 140 further uses the prediction image sequence and the prediction / non-prediction information in addition to the existing APD scheme in generating the difference sequence, thereby further improving the frequency at the peak point, A differential sequence into which data can be inserted can be generated.
Each element of the difference sequence D is denoted by D _i , The difference sequence generation unit 140 can generate a difference sequence using the elements of the cover image sequence and the elements of the predicted image sequence as shown in Equation (6) below.

Then, the histogram generator 150 of the data concealment apparatus 100 generates a histogram corresponding to the difference sequence. Here, the histogram generation is for applying the data concealment technique (APD) using differential sequence shift to obtain shift information necessary for shifting the differential sequence, and its technical structure is described in YC Li, CM Yeh, and CC Chang, Data hiding based on the similarity between neighboring pixels with reversibility, Digital Signal Processing, Vol. 20, No. 4, pp 1116-1128, July 2010.
The histogram generator 150 generates a histogram of the difference sequence and outputs information for shifting the difference sequence, that is, a peak point PP having the maximum frequency _One , peak point 1), the peak point having the second highest frequency (PP ₂ , peak point 2), PP _One CZP &lt; / RTI &gt; having a frequency value closest to &lt; RTI ID = 0.0 & _One (closest zero point 1), PP ₂ Lt; RTI ID = 0.0 &gt; CZP &lt; / RTI &gt; ₂ (closest zero point 2).
Accordingly, the confidential data inserting unit 160 can generate the differential sequence DS shifted by shifting the differential sequence using the information of the histogram according to the following Equation (7).

The confidential data inserting unit 160 may insert the confidential data into the DS according to an arithmetic process as shown in Equation (8) below to generate the embedding sequence DE.

The hidden image sequence generating unit 170 generates an image sequence S for the cover image sequence C and the embedding sequence DE to finally generate the hidden image sequence S, thereby obtaining each peak point of the histogram for the difference sequence It is possible to insert the confidential data into the pixels corresponding to each bit by bit.

Accordingly, the hidden image output unit 180 may scan the hidden image sequence to construct a hidden image in an inverse s-order, and then output the hidden image.
FIGS. 5 and 7 to 8 are illustrations showing a cover image, a predictive image generated from a cover image, and a hidden image embedded with confidential data.
In FIGS. 5 and 7 to 8, confidential data embedding is exemplarily shown by the entire image processing.
First, in Figs. 5 and 7 to 8, the coefficient alpha used in Equation 1 and Equation 2 at the time of generating the predictive image is set to 0.298, and the threshold value used in Equation 4 is set to 15 Of the predicted image. Accordingly, the first temporary predictive pixel value V (i) at each position from (1) to (8) in Fig. 5 _One 157.6, 157.7, 157.5, 158.0, 157.6, 158.7, and 158.2, respectively. Then, the calculated first provisional predicted pixel value V1 is applied to Equation (2) to obtain a deviation (dev), which is 10.9, 10.3, 8.1, 7.3, 13.2, 8.4, 15.3, 12.2.
Accordingly, if the threshold value beta is set to 15, the deviation (15.3) at the position 7 is equal to or larger than the threshold value beta as shown in FIGS. 5 and 7 to 8, It is possible to generate the predicted image by copying the pixel value 159 at the corresponding position in the cover image without further performing the pixel value predicting process by discriminating the position ⑦ as a place with low spatial locality.
The predicted image generating unit 130 estimates the surface characteristics of the image in the remaining positions 1 to 6 and 8 in which the deviation is less than the threshold value beta to obtain a simple inclined surface or a flat surface The estimated positions ①, ④, ⑥ can be distinguished.
Then, the predictive image generation unit 130 generates the predicted image data of the second temporary predicted pixel value V (i, j, ₂ (157), 158.0, and 157.0 are calculated according to Equation (4), and the predicted image values are calculated by rounding this value (the second temporary predicted pixel value) to a predicted pixel value at each position .
On the other hand, since the predicted image generating unit 130 estimates that the surface characteristics at the other positions?,?,?,? Are not a simple slope or a plane, the first temporary predictive pixel value V _One The predicted pixel value at the corresponding position may be determined and a predicted image may be generated.
A prediction image constructed through such a process can be described with reference to FIGS. 5, 7 to 8. FIG. As shown in FIG. 8, when the difference sequence D according to the embodiment of the present invention is generated and the histogram is generated, _One = 0, PP ₂ = 1 and h (PP _One ) = 12, h (PP ₂ ) = 3 can be generated. Accordingly, the maximum number of data bits that can be embedded in the difference sequence and concealable in the cover image is h (PP _One ) And h (PP ₂ ), And it can be confirmed that it is increased by 4 bits or more as compared with the above-described APD method. If the threshold value? Is set to 20, the predicted pixel value is also used at the position? _One ) = 13, h (PP ₂ ) = 3, and the number of bits of confidential data to be inserted can be increased by one bit. In the embodiment of the present invention, if the threshold value beta is set to 0, all the pixels of the predicted image are values obtained by copying the pixel values of the cover image as they are. Therefore, It can operate in the same manner as the APD technique.
FIGS. 9 to 10 are flowcharts for explaining a data restoration method according to an embodiment of the present invention, and are illustrations for explaining a process thereof.

9, the data decompression apparatus 200 according to the embodiment of the present invention receives a concealed image through the concealed image receiving unit 210 and converts the received concealed image from left to right, from top to bottom And scans in an inverse s-order to generate a concealed image sequence S composed of pixel values (S201).
The prediction-based extraction mode determination unit 220 of the data decompression apparatus 200 estimates spatial regions and surface characteristics for each position of the hidden image sequence to determine whether to use the predicted pixel values, And determines the extraction mode (S203).
The prediction-based extraction mode determination unit 220 determines a corresponding extraction mode among the three extraction modes at the position of each element of the hidden image sequence. That is, the prediction-based extraction mode determination unit 220 determines the prediction mode based on i (i If the prediction pixel value is not used in both the previous position and the current position, the extraction mode is determined as Mode 1. If the prediction pixel value is used in the current position, the extraction mode is determined as Mode 2 , And if the prediction pixel value is used in the previous position but the prediction pixel value is not used in the current position, the extraction mode is determined as Mode 3.
The operation of the prediction-based extraction mode determination unit 220 will be described in detail as follows. The prediction-based extraction mode determination unit 220 can determine the extraction mode at each element position of the concealed image sequence S in FIG.
If the current position is i, the embedding sequence is DE _i And the cover image sequence is S _i-1 The restoration was completed. Detailed procedures for restoring the embedding sequence and the cover image sequence will be described later.
First, the prediction-based extraction mode determination unit 220 determines whether the location i is a pixel value predictable area. If it is not a pixel value predictable area, the prediction-based extraction mode determination unit 220 can determine that the predicted pixel value is not used at the corresponding position i.
If the current position is a pixel value predictable area, the prediction-based extraction mode determination unit 220 uses the adjacent pixel values of the restored cover image to calculate a restored first temporary predictive pixel value based on the spatial locality (Dev) between the reconstructed first temporary predictive pixel value and the adjacent pixel values is calculated in the same manner as in Equation (2) so as to determine the degree of spatial locality, and the calculated deviation dev is compared with the threshold value beta) or more.
The prediction-based extraction mode determination unit 220 can determine that the predicted pixel value is not used at the corresponding position i because the spatial locality is low if the calculated deviation dev is equal to or greater than the threshold value beta. If the calculated deviation dev is less than the threshold value?, The prediction-based extraction mode determination unit 220 may determine that the predicted pixel value is used because the spatial locality is high. The prediction-based extraction mode determination unit 220 may repeat this process while increasing the position i value to determine the extraction mode according to whether pixel values are used at the previous position and the current position.
If the prediction pixel value is used at the current position and the extraction mode is Mode 2, the prediction-based extraction mode determination unit 220 determines the prediction prediction pixel value RP _i ), The surface characteristics are estimated in the same manner as in the hidden method. The prediction-based extraction mode determination unit 220 determines the restored second temporary predicted pixel value using the adjacent pixel values in the previously restored cover image when the estimated surface characteristic is a simple inclined surface or a flat surface (4), and the restored predicted pixel value RP _i ). If the estimated surface characteristic is not a simple slope or plane, the prediction-based extraction mode determination unit 220 determines the restored first temporary predicted pixel value as the restored prediction pixel value RP _i ).
If the extraction mode is Mode 2 or Mode 3, the prediction-based extraction mode determination unit 220 determines that the associated value AV _i (Associated Value at i ^th position can be calculated using the following equation (12).
In addition, the embedding sequence restoring unit 230 and the cover image sequence restoring unit 240 of the data restoring apparatus 200 restore the embedding sequence from the hidden image sequence, extract the cover image from the hidden image sequence according to the extraction mode determined at each position, The sequence is restored (S205).
The secret data extracting unit 250 extracts confidential data from the embedding sequence and outputs it (S207). The extracted confidential data is completely restored without loss and becomes the same as confidential confidential data.
The cover image output unit 260 scans the cover image sequence to construct a cover image in inverse s-order, and then outputs the restored cover image (S209).
More specific operations and implementation methods of the prediction-based extraction mode determining unit 220, the embedded sequence restoring unit 230, the cover image sequence restoring unit 240, and the classified data extracting unit 250 will be described below.
The embedding sequence restoring unit 230 and the cover image sequence restoring unit 240 repeatedly increase the position i value in the hidden image sequence from 0 to imagewidth * imageheight-1, If you do, you can restore the embedding sequence DE and the cover image sequence C in Figure 10.
The embedding sequence restoring unit 230 and the cover image sequence restoring unit 240 at the position i can perform the following processing from the corresponding position i according to the extraction modes (Mode 1 to Mode 3) determined by the prediction-based extraction mode determining unit 220, When the process of each mode is performed, the embedding sequence DE _i And cover image sequence C _i Can be restored.
Mode 1: Extraction mode when the top two rows of the image (i≤imagewidth * 2) or the prediction pixel value is not used in both the previous position and the current position. In Mode 1, the embedding sequence DEi and the cover image sequence C (i) are calculated by using the following Equations (10) and (11) in the same manner as in the APD (Adjacent Pixel Difference) _i Can be restored.

Mode 2: Extraction mode when the predicted pixel value is used at the current position. Using the above equation (10), the embedding sequence DE _i And uses the following equation (12) to calculate the cover image sequence C _i . RP _i Is the restored predicted pixel value calculated at the position i when the extraction mode determination result at the position i is Mode 2 in the restoration process, and can be calculated in the same manner as the hidden process. In addition, _i (Associated Value at i ^th position is a value calculated as Equation (12) when the extraction mode determination result is Mode 2 or Mode 3 at the position i during the restoration process, and is used for restoring the cover image. .

Mode 3: Extraction mode when the predicted pixel value is used in the previous position but the predicted pixel value is not used in the current position. Using the equation 10, the embedding sequence DE _i And uses the following equation (13) to calculate the cover image sequence C _i . RP _i-1 Is the restored predicted pixel value calculated at the position i-1 when the extraction mode determination result at the position i-1 is Mode 2 in the restoration process. AV _i Is the value calculated in the above equation (12).

More specifically, the confidential data extracting unit 250 extracts confidential data hidden in the embedding sequence by applying Equation (14) to the embedding sequence DE (S207). The extracted confidential data is the same as the confidential confidential data because it is completely restored without loss.

Then, the restoration apparatus 200 scans the restored cover image sequence through the cover image output unit 260, restores the cover image in inverse s-order, and outputs the restored cover image ( S209).
The restoration process as described above is specifically illustrated in FIG. As shown in Fig. 10, the embedding sequence DE and the cover image sequence C can be restored from the hidden image sequence S. When the confidential data hidden from the embedding sequence is extracted, the original confidential data "101011110110101" . The extracted confidential data coincides with the confidential confidential data.
As described above, according to the data concealment method using the differential sequence shift, pixel value prediction is precisely performed according to the surface characteristics of a region having high spatial localities according to the embodiment of the present invention, The frequency at the peak point in the histogram of the sequence greatly increases, so that the number of bits of insertable confidential data can be greatly increased. In addition, it is possible to provide an operation method capable of concealing confidential data at various levels while keeping the image quality of the concealed image so excellent as to be difficult to distinguish from the cover image according to the adjustment of the threshold value?. In addition, the confidential data and the original cover image can be completely restored from the hidden image without loss, and can be applied effectively to medical, defense, and digital library applications.
Meanwhile, the above-described methods according to the present invention may be stored in a computer-readable recording medium. The computer-readable recording medium may be a ROM, a RAM, a CD-ROM, Magnetic tapes, floppy disks, optical data storage devices, and the like.
The computer readable recording medium may be distributed over a networked computer system so that computer readable code can be stored and executed in a distributed manner. And, functional programs, codes, and code segments for implementing the above method can be easily deduced by programmers of the technical field to which the present invention belongs.
While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, It will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the present invention.

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

100: Data hiding device
200: data restoration device

Claims (14)

In the data concealment method,
Generating a predicted image based on spatial locality and surface characteristics from the cover image;
Generating a difference sequence using a predicted image sequence obtained from the predicted image and a cover image sequence obtained from the cover image;
Generating a histogram corresponding to the difference sequence;
Generating a shifted difference sequence by shifting the differential sequence according to shift information obtained from the histogram, and embedding confidential data in the shifted difference sequence;
Generating a hidden image sequence by combining the embedded sequence and the cover image sequence according to the embedding; And
And outputting a concealed image according to the concealed image sequence,
When the hidden image is scanned by the restoration device to generate a hidden image sequence, the extraction mode is determined and restored according to whether the spatial locality based prediction value is used for each element of the hidden image sequence generated by the restoration device,
The restoration device identifies whether a predicted pixel value can be calculated for a current hidden image sequence, identifies whether a predicted value is used according to the possibility, determines an extraction mode corresponding to the current hidden image sequence doing
A reversible data concealment method applying pixel value prediction technique. The method according to claim 1,
Wherein the generating the prediction image comprises:
And generating a first temporary predictive pixel value based on spatial locality from pixel information of a predetermined region in the cover image
A reversible data concealment method applying pixel value prediction technique. 3. The method of claim 2,
Wherein the generating the prediction image comprises:
Determining a spatial locality from a deviation between the first temporary predicted pixel value and surrounding pixel values;
Determining a surface characteristic for the predetermined region when the deviation is less than a threshold value;
Calculating a second provisional predicted pixel value according to the surface property; And
And generating the predictive image by selectively applying the first temporary predictive pixel value, the second temporary predictive pixel value, and the pixel value of the cover image according to the spatial locality and the surface characteristic,
A reversible data concealment method applying pixel value prediction technique. The method according to claim 1,
Wherein the step of generating the difference sequence comprises:
Generating a difference sequence selectively applying a cover image sequence and a predictive image sequence according to whether the predictive pixel value is used for the cover image sequence and the predictive image sequence,
A reversible data concealment method applying pixel value prediction technique. The method according to claim 1,
Wherein the embedding comprises:
Shifting the difference sequence according to shift information obtained from the histogram and inserting the confidential data at a peak point of the shifted difference sequence
A reversible data concealment method applying pixel value prediction technique. The method according to claim 1,
Wherein the confidential data includes watermark data including owner information or original confirmation information of the cover image
A reversible data concealment method applying pixel value prediction technique. In a data concealment apparatus,
A predicted image generating unit that generates a predicted image based on spatial locality and surface characteristics from a cover image;
A difference sequence generation unit for generating a difference sequence by using a predicted image sequence obtained from the predicted image and a cover image sequence obtained from the cover image;
A histogram generator for generating a histogram corresponding to the difference sequence;
A confidential data inserter for shifting the difference sequence according to information of the histogram to generate a shifted difference sequence and embedding confidential data in the shifted difference sequence;
A hidden image sequence generator for generating a hidden image sequence by combining the embedded sequence and the cover image sequence according to the embedding; And
And a hidden image output unit according to the hidden image sequence,
When the hidden image is scanned by the restoration device to generate a hidden image sequence, the extraction mode is determined and restored according to whether the spatial locality based prediction value is used for each element of the hidden image sequence generated by the restoration device,
The restoration device identifies whether a predicted pixel value can be calculated for a current hidden image sequence, identifies whether a predicted value is used according to the possibility, determines an extraction mode corresponding to the current hidden image sequence doing
A reversible data concealment device applying pixel value prediction technique. 8. The method of claim 7,
The predictive image generator may include:
Generating a first temporary predicted pixel value based on spatial locality from pixel values of the predetermined region of the cover image; determining spatial localities from the deviation between the first temporary predicted pixel value and its surrounding pixel values; the second temporary predictive pixel value is calculated according to the surface characteristic, and the first temporary predictive pixel value, the second temporary predictive pixel value, and the second temporary predictive pixel value are calculated according to the spatial locality and the surface characteristic, And selectively applying the second temporary predictive pixel value and the pixel value of the cover image to generate the predictive image
A reversible data concealment device applying pixel value prediction technique. 8. The method of claim 7,
Wherein the difference sequence generation unit selectively applies the cover image sequence and the predicted image sequence according to whether a predicted pixel value is used or not to generate a difference sequence composed of pixel value differences,
Wherein the confidential data inserting unit shifts the difference sequence according to shift information obtained from the histogram and inserts the confidential data at a peak point of the shifted difference sequence
A reversible data concealment device applying pixel value prediction technique. 8. The method of claim 7,
Wherein the confidential data includes watermark data including owner information or original confirmation information of the cover image
A reversible data concealment device applying pixel value prediction technique. A method for restoring concealed data,
Generating a hidden image sequence by scanning a hidden image;
Determining an extraction mode according to spatial locality and whether a surface property-based prediction value is used at a position of each element of the hidden image sequence;
Restoring the embedding sequence from the hidden image sequence, and recovering the cover image sequence from the hidden image sequence according to the determined extraction mode; And
Acquiring and outputting confidential data extracted or restored from the embedding sequence,
Wherein the step of determining the extraction mode comprises:
Identifying whether a pixel value is predictable at each position of the hidden image sequence; And
Identifying whether to use the prediction value according to the possibility, and determining an extraction mode corresponding to each position of the hidden image sequence
A method for restoring hidden data. delete 12. The method of claim 11,
Wherein,
Determining a first mode (Mode 1) if the current position is the upper two rows of the image, or if the predicted pixel value is not used in both the previous position and the current position, and restoring according to the APD (Adjacent Pixel Difference)
(Mode 2) when the predicted pixel value is used at the current position, and determines the restored predicted pixel value (RP _i ) at the current position calculated in the mode determination process and the associated value AV _i ), &_lt; / RTI &gt; And
If the predicted pixel value is used in the previous position but the predicted pixel value is not used in the current position, it is determined as the third mode (Mode 3), and the restored predicted pixel value (RP _i-1 ) and the associated value (AV _i ) at the current position
A method for restoring hidden data. A computer program stored on a medium for execution in a computer, the method according to any one of claims 1 to 6, 11 and 13.

Images