Figure 1.
Example of the cross division.
Citation: Megan K. Creutzburg, Emilie B. Henderson, David R. Conklin. Climate change and land management impact rangeland condition and sage-grouse habitat in southeastern Oregon[J]. AIMS Environmental Science, 2015, 2(2): 203-236. doi: 10.3934/environsci.2015.2.203
| [1] | Rong Li, Xiangyang Li, Yan Xiong, An Jiang, David Lee . An IPVO-based reversible data hiding scheme using floating predictors. Mathematical Biosciences and Engineering, 2019, 16(5): 5324-5345. doi: 10.3934/mbe.2019266 |
| [2] | Dingwei Tan, Yuliang Lu, Xuehu Yan, Lintao Liu, Longlong Li . High capacity reversible data hiding in MP3 based on Huffman table transformation. Mathematical Biosciences and Engineering, 2019, 16(4): 3183-3194. doi: 10.3934/mbe.2019158 |
| [3] | Yan Li, Junwei Wang, Shuangkui Ge, Xiangyang Luo, Bo Wang . A reversible database watermarking method with low distortion. Mathematical Biosciences and Engineering, 2019, 16(5): 4053-4068. doi: 10.3934/mbe.2019200 |
| [4] | Yan-Xiao Liu , Ching-Nung Yang, Qin-Dong Sun . Enhance embedding capacity for switch map based multi-group EMD data hiding. Mathematical Biosciences and Engineering, 2019, 16(5): 3382-3392. doi: 10.3934/mbe.2019169 |
| [5] | Yuzhen Zhou, Erxi Zhu, Shan Li . An image encryption algorithm based on the double time-delay Lorenz system. Mathematical Biosciences and Engineering, 2023, 20(10): 18491-18522. doi: 10.3934/mbe.2023821 |
| [6] | Guodong Ye, Huishan Wu, Kaixin Jiao, Duan Mei . Asymmetric image encryption scheme based on the Quantum logistic map and cyclic modulo diffusion. Mathematical Biosciences and Engineering, 2021, 18(5): 5427-5448. doi: 10.3934/mbe.2021275 |
| [7] | Cheonshik Kim, Dongkyoo Shin, Ching-Nung Yang . High capacity data hiding with absolute moment block truncation coding image based on interpolation. Mathematical Biosciences and Engineering, 2020, 17(1): 160-178. doi: 10.3934/mbe.2020009 |
| [8] | Xianyi Chen, Anqi Qiu, Xingming Sun, Shuai Wang, Guo Wei . A high-capacity coverless image steganography method based on double-level index and block matching. Mathematical Biosciences and Engineering, 2019, 16(5): 4708-4722. doi: 10.3934/mbe.2019236 |
| [9] | Li Li, Min He, Shanqing Zhang, Ting Luo, Chin-Chen Chang . AMBTC based high payload data hiding with modulo-2 operation and Hamming code. Mathematical Biosciences and Engineering, 2019, 16(6): 7934-7949. doi: 10.3934/mbe.2019399 |
| [10] | Zilong Liu, Jingbing Li, Jing Liu . Encrypted face recognition algorithm based on Ridgelet-DCT transform and THM chaos. Mathematical Biosciences and Engineering, 2022, 19(2): 1373-1387. doi: 10.3934/mbe.2022063 |
Due to the extensive use of digital image, several technologies have been developed to protect the copyright and the integrity of the digital images, including image hashing [1] and data hiding. Data hiding is a technology that embeds additional information imperceptibly into the host media of various formats [2,3,4]. The embedded data can be extracted in a particular way for authentication or other purposes.
Reversible data hiding (RDH) is a sort of data hiding technology that makes it possible to reverse the embedding of the secret data, and the host media can be recovered after the data have been embedded. Recently, many RDH methods, such as difference expansion [5,6,7], histogram shifting [8,9,10], and pixel value ordering [11,12,13], have been developed for reversibly hiding data in plaintext images. By using the redundancy and spatial correlation, these methods can embed additional information into plaintext images with fully reversibility. For the encrypted image, the redundancy and spatial correlation are lost. So the plaintext image-based methods cannot be used in the encrypted domain.
With the development of cloud services, RDH in encrypted images (RDHEI) has been valued. Since cloud services are used extensively now in many different applications, continually increasing quantities of data are being stored and processed in the cloud instead of in the content owners' terminals. For privacy, the content owner always provides encrypted images instead of plaintext images. Therefore, it is a problem for the cloud service provider to process the encrypted image for reversible data hiding when the content of the image cannot be accessed.
To date, researchers have proposed several RDHEI methods that use different ways to embed additional data reversibly in the encrypted domain. Some methods perform bit flipping on the bit-wise encrypted image to embed additional data. To extract the embedded data and recover the image, the flipped bits are detected in the decrypted image depending on the spatial correlation in the decrypted domain. In [14], a 3-LSB (least significant bit) flipping scheme was designed to embed one additional bit into one block of the encrypted image. After decrypting the image, the flipped pixels in each block were detected by a designed smoothness estimator so the embedded bits could be extracted and the image could be recovered. The improved methods of [14] were proposed to be used in [15,16,17]. In [15], a side-match technology was used jointly with the smoothness estimator to enhance the reversibility. In [16], to improve the accuracy of image recovery, the author used a more precise block-smoothness estimator, and the blocks were recovered in descending order of their smoothness. In [17], the way in which data were embedded was improved for high visual quality, and an image content-based estimator was used to extract the data and recover the image. In [18], the encrypted image was divided randomly into same-size groups, and one additional bit was hidden in one group by flipping one bit of each pixel of the group. After the decryption of the image, the flipped pixels were detected by using prediction error.
Some methods use compression of the images to make additional room available in the encrypted image to accommodate the secret data. In [19], a matrix compression scheme was used to compress the LSBs of the encrypted image to accommodate additional data. In [20], the encrypted image was decomposed into blocks, and the blocks were classified as smooth blocks or complex blocks. A compression matrix was used to compress the LSBs of the smooth blocks to accommodate secret bits. In [21], according to the compression effect, one of the two compression schemes, which were a run-length coding scheme and a matrix compression scheme, was chosen to compress the blocks of the encrypted image respectively. In [22], by using the LDPC code for compression, the additional data can be placed in the fourth LSBs of the encrypted image. In [23], the MSBs of the encrypted image was encoded by using LDPC code to place additional bits.
Some methods partially maintain the spatial correlation in encrypted images by using a customized encryption schemes, e.g., the traditional RDH schemes, such as histogram shifting (HF) and pixel value ordering (PVO). In [24], the encrypted image was decomposed into cross blocks, and pixels in the same cross block were encrypted with the same encryption key by using the additive hormomorphic cryptosystem. Based on the encrypted cross blocks, the difference histogram was generated and a designed histogram shifting scheme was used for data embedding. In [25] and [26], the original images were encrypted at the level of $2 \times 2$ blocks. In [25], the additional data can be embedded into blocks by using an extended PVO scheme. In [26], a set of prediction errors is generated by using a block-level predictor, and an extended prediction error expansion scheme was designed to embed secret data.
In this paper, we propose a new RDHEI method that uses two-phase histogram shifting. By using a designed image division strategy and the RC4 cryptosystem, the method partially maintains the spatial correlation in the encrypted image. To achieve high embedding capacity, the process of data embedding consists of two embedding phases. In each embedding phase, a difference histogram can be obtained from the encrypted image; then, a designed histogram shifting scheme is used for data embedding based on the difference histogram. In summary, the merits of this paper are:
(1) Based on the division of the image and shifting the two-phase histogram, the proposed method can provide a high capacity for data embedding.
(2) The recovery of the image and the extraction of data can be performed in the encrypted domain or in the decrypted domain.
(3) The recovery of the image and the extraction of data are error-free.
The rest of the paper is organized as follows. In Section Ⅱ, the homomorphic cryptosystem and Li et al.'s method are described in detail. In Section Ⅲ, the proposed method is introduced. Section Ⅳ presents the experimental results and comparisons, Section 5 gives the conclusion of this paper.
In 1978, Rivest et al. introduced the idea of hormomorphic encryption [27], and it commonly is defined as follows. If m1 and m2 are two words in the set of the plaintexts, for the encryption function E and any given encryption key, there is:
| $ E({m_1}{ \odot _P}{m_2}) \leftarrow E({m_1}){ \odot _E}E({m_2}) $ | (1) |
where ${ \odot _P}$ is a certain operator for plaintexts ${m_1}$ and ${m_2}$, ${ \odot _E}$ is a certain operator for ciphertexts $E({m_1})$ and $E({m_2})$, and $ \leftarrow $ is "directly calculated from." By using the homomorphism, some special calculations can be performed on encrypted data without decryption.
The stream cipher RC4 is one of the well-known ciphers that are used for additive homomorphic encryption. If m and c denote the plaintext and the ciphertext, respectively, and k denotes the pseudo-random encryption key generated from a secret seed by using RC4, the encryption and decryption are performed as follows:
| $ {\rm Encryption:} ~c = E(m, k) = (m + k){\rm mod} ~n $ | (2) |
| $ {\rm Decryption}: ~m = D(c, k) = (c - k){\rm mod}~ n $ | (3) |
where n is a preset modulus. The encryption in Eq. (2) has the property of additive homomorphism. Noting that m1 and m2 are two plain values and that k1 and k2 are two corresponding encryption keys, then:
| $ (E(m1,k1)+E(m2,k2))modn=((m1+k1)modn+(m2+k2)modn)modn =(m1+m2+k1+k2)modn =E(m1+m2,k1+k2) $ | (4) |
| $ D((E(m1,k1)+E(m2,k2))modn,k1+k2)=D(E(m1+m2,k1+k2),k1+k2) =(m1+m2)modn $ | (5) |
| $ (E({m_1}, {k_1}) + {m_2})\bmod n = E({m_1} + {m_2}, {k_1}) $ | (6) |
| $ D((E({m_1}, {k_1}) + {m_2})\bmod n, {k_1}) = ({m_1} + {m_2})\bmod n $ | (7) |
According to Eqs. (6) and (7), by using the property of homomorphism, the modular n addition operation can be performed on the encrypted data to modify the plain data without decryption.
Li et al. proposed an RDHEI method using cross division and additive homomorphism [24]. Due to additive homomorphism, data extraction can be performed in both the encrypted domain and the decrypted domain.
Before encryption, cross division is used to divide the original image into cross-blocks. Each cross block consists of a 'central pixel' and all of the pixels that are adjacent to the central pixel. For an image, I, sized $M \times N$, the (x, y) coordinates of the central pixels are defined as:
| $ \left\{ {x=xy=2xmod5+5z} \right.{\text{, where }}x{\text{ = 1, 2, }}...{\text{, }}M{\text{, }}z{\text{ = 0, 1, }}...{\text{, }}\left\lfloor {{{(N - 2x\bmod 5)} / 5}} \right\rfloor $ | (8) |
Figure 1 shows an example of cross division. The black pixels are the central pixels of the cross-blocks, and the white pixels adjacent to the central pixels belong to the same cross-blocks. The gray pixels do not belong to any cross-block.
To encrypt the image I, the RC4 stream cipher is used to generate an encrypted matrix, K, with the size of $M \times N$. The encryption scheme is described as:
| $ {I_E} = E(I, K) = (I + K)\bmod256 = ({p_{i, j}} + {k_{i, j}})\bmod256{\text{, }}i = 1, 2, ..., M{\text{, }}j = 1, 2, ..., N $ | (9) |
To keep the same difference between the neighboring pixels in the cross blocks, the encrypted matrix K is generated as follows: if ${p_{i1, j1}}$ and ${p_{i2, j2}}$ belong to the same cross-block, they should be encrypted by the same key, i.e., ${k_{i1, j1}} = {k_{i2, j2}}$.
When the data hider receives the encrypted image, a difference histogram is generated from all of the cross-blocks of the encrypted image. For each cross-block with the encrypted central pixel ${e_{i, j}}$, the differences of the cross-block are generated as follows:
| $ \left\{ {di−1,j=(ei−1,j−ei,j)mod256,when i⩾2di+1,j=(ei+1,j−ei,j)mod256,when i⩽M−1di,j−1=(ei,j−1−ei,j)mod256,when j⩾2di,j+1=(ei,j+1−ei,j)mod256,when j⩽N−1} \right. $ | (10) |
Note that the four differences are identical to the difference generated from the plain image. For example, ${d_{i - 1, j}}$ is computed from ${e_{i - 1, j}}$ and ${e_{i, j}}$, where ${e_{i - 1, j}} = ({p_{i - 1, j}} + {k_{i - 1, j}})\bmod256$, ${e_{i, j}} = ({p_{i, j}} + {k_{i, j}})\bmod256$, and ${k_{i - 1, j}} = {k_{i, j}}$, we have
| $ di−1,j=(ei−1,j−ei,j)mod256 =((pi−1,j+ki−1,j)mod256−(pi,j+ki,j)mod256)mod256 =(pi−1,j−pi,j)mod256 $ | (11) |
Due to the spatial correlation, the neighboring pixels are likely to be similar. Therefore, most of the differences are close to 0 and 255.
After the differences of all of the cross-blocks were generated, the difference histogram was formed from all of the differences. Figure 2 shows an example of the difference histogram of the encrypted image, Lena (Figure 8(d)). The figure shows that most of the differences were close to the two sides.
Based on the difference histogram, a variant histogram shifting scheme is used for data embedding. The variant histogram shifting scheme shifts the bins from the two side to the middle (bin 127 and bin 128) to vacate room for embedding. The procedure of data embedding is given as follows:
Step 1: Shift bins 1–126 one step to their right neighboring bins, i.e., bin 1 is shifted to bin 2, bin 2 is shifted to bin 3, …, bin 126 is shifted to bin 127. The bins are shifted by performing modular 256 addition on the corresponding encrypted pixels. For example, if ${d_{i - 1, j}} = 1$ is one element of bin 1, ${d_{i - 1, j}}$ can be shifted to bin 2 as follows:
| $ di−1,j+1=((ei−1,j−ei,j)mod256+1)mod256 =((ei−1,j+1)mod256−ei,j)mod256 $ | (12) |
That is, the shifting can be performed by modifying ${e_{i - 1, j}}$ to $({e_{i - 1, j}} + 1)\bmod256$.
Then, bins 129–254 are shifted one step to the left neighboring bins, i.e., 128–253. The bins are shifted by performing modular 256 subtraction on the corresponding encrypted pixels. The processing is similar to the shifting of bins 1-126. For example, to shift the element ${d_{i - 1, j}} = 254$ to bin 253, ${e_{i - 1, j}}$ is modified to $({e_{i - 1, j}} - 1)\bmod 256$.
Step 2: A boundary map is generated to distinguish the new elements from the original elements in bins 127 and 128. The boundary map is a binary sequence. For all encrypted pixels that correspond to the elements in bins 127 and 128, the original pixels are denoted by 0 and the modified pixels are denoted by 1 in the boundary map.
Step 3: After Steps 1 and 2, bins 1 and 254 are empty. Then, bins 0 and 255 can be used for embeddding data. The length of the secret data is embedded first, and, then, the secret data and the boundary map are embedded. To embed a bit 1, the element in bin 0 is shifted to bin 1, or the element in bin 255 is shifted to bin 254. To embed a bit 0, do nothing. The processing of element shifting is the same as that in Step 1.
Step 4: Steps 1–3 can be performed many times until all of the secret data have been embedded. In the second round, bins 3–126 are shifted step to the right to bins 4–127, and bins 129–252 are shifted one step to the left to bins 128–251. Then, bin 2 and bin 253 are used for embedding data. In the third round, bins 5–126 and bins 129–250 are shifted, and bins 4 and 251 are used for embedding data. The boundary map should be generated and embedded in each round.
When the receiver obtains the marked encrypted image, data extraction and image recovery can be performed in two ways: (1) Extract the secret data and recover the image in the encrypted domain; then, decrypt the encrypted image; (2) Perform image decryption first, then extract the embedded data and recover the image in the decrypted domain.
(1) Processing in the encrypted domain: Since the generation of the difference histogram and embedding data are processed directly on the encrypted image, data extraction and image recovery can be performed directly in the encrypted domain as follows:
Step 1: Generate the difference histogram from the marked encrypted image.
Step 2: Extract the length of the secret data from the elements in bin 0 and bin 255.
Step 3: Compute the round number according to the length of the secret data and the bins of the difference histogram.
Step 4: The reverse processing of data embedding is performed from the last round to the first round. In each round, first, the boundary map and the embedded data of the round are extracted, and the two bins used for embedding data are recovered; then, with the aid of the boundary map, the shifted bins are recovered to their initial positions in this round.
Step 5: Reorder all of the extracted data to obtain the secret data, and decrypt the recovered encrypted image to obtain the original image.
(2) Processing in the decrypted domain: The marked encrypted image can be directly decrypted to a marked decrypted image as follows:
| $ Im=E(Ime,K)=(Ime−K)mod256=(emi,j−ki,j)mod256, i=1,2,...,M, j=1,2,...,N $ | (13) |
The difference histogram can be generated from the decrypted image for data extraction and image recovery. Due to additive homomorphism, performing the modular 256 addition performed on the encrypted pixels is also performing that on the decrypted pixels. Taking ${e_{i - 1, j}}$ as an example, according to Eqs. (6) and (7), if $e_{i - 1, j}^m = (e_{i - 1, j}^{} + s)\bmod256$, then $e_{i - 1, j}^m$ is decrypted as:
| $ p_{i - 1, j}^m = (p_{i - 1, j}^{} + s)\bmod 256 $ | (14) |
According to Eq. (11),
| $ {d_{i - 1, j}} = (e_{i - 1, j}^m - e_{i, j}^{})\bmod256 = (p_{i - 1, j}^m - p_{i, j}^{})\bmod256 $ | (15) |
Eq. (14) proves that the modification of the encrypted pixel is the same as that of the decrypted pixel, and Eq. (15) proves that the difference histogram of the marked decrypted image is the same as that of the marked encrypted image. Therefore, data extraction and image recovery in the decrypted domain are the same as they are in the encrypted domain, and the recovered image is identical to the original image.
To improve the embedding capacity of Li et al.'s method [24], in this section, we propose a new RDHEI method based on two-phase histogram shifting. An overview of the proposed method is given in Figure 3. First, the image is decomposed into non-overlapping T-type blocks and encrypted. Then, the data hider performs two-phase histogram shifting for data embedding. In each phase, a difference histogram is generated for embedding data. Finally, the receiver can perform data extraction and image recovery in the encrypted domain or in the decrypted domain.
We propose a T-type division scheme to divide the image into T-type blocks. Figure 4 shows an example of T-type division in which each T-type block consists of four pixels, i.e., the central pixel, which is in the center of the block (marked as 'c'); the up pixel, which is above the central pixel (marked as 'u'); the down pixel, which is below the central pixel (marked as 'd'); and the side pixel, which is to the left or right of the central pixel (marked as 's').
For an image I sized $M \times N$, the coordinates (x, y) of the central pixel of each of the T-type blocks are defined as:
| $ \left\{ {x=2,4,6,.....,2×⌊M−12⌋y=(x2+1)mod2+z, z=1,3,5,...,2×⌊N2⌋−1} \right. $ | (16) |
For each block with the central pixel (x, y), the coordinate (sx, sy) of the side pixel are defined as:
| $ \left\{ sx=xsy={y+1, if (x2+1)mod2=0y−1, if (x2+1)mod2=1 \right. $ | (17) |
After the image is divided, a pseudo-random matrix K is generated by using the RC4 steam cipher. Then, image, I, is encrypted by the matrix K, as follows:
| $ IE=E(I,K)=(I+K)mod256=(pi,j+ki,j)mod256, i=1,2,...,M, j=1,2,...,N $ | (18) |
To generate the difference histogram, pixels in the same T-type block should be encrypted by using the same encryption key. For example, if a T-type block consists of four pixels, i.e., ${p_{i, j}}$, ${p_{i - 1, j}}$, ${p_{i + 1, j}}$, and ${p_{i, j - 1}}$, the four corresponding encryption keys, ${k_{i, j}}$, ${k_{i - 1, j}}$, ${k_{i + 1, j}}$, and ${k_{i, j - 1}}$ should be identical. After encryption, the content owner send the encrypted image for data embedding.
The process of embedding data consists of two phases. In the first phase, for each T-type block, the difference of the central pixel and one of its three neighboring pixels is computed. A 'small' difference histogram is formed with these differences, and difference histogram shifting is performed for a limited number of rounds to embed the secret data. The elements in the small difference histogram are shifted by modifying the values of the central pixels. In the second phase, for each T block, three differences are computed, i.e., the difference of the up pixel and the central pixel, the difference of the down pixel and the central pixel, and the difference of side pixel and the central pixel. A 'large' difference histogram is formed with these differences, and difference histogram shifting is performed for embedding data. The elements in the large difference histogram are shifted by modifying the values of the up pixels, the down pixels, and the side pixels.
In the first phase, the central pixels are used for embedding data. The process of embedding data in the first phase is proposed as follows:
Step 1: For each encrypted T-type block, denote the central pixel, the up pixel, the down pixel, and the side pixel as ec, eu, ed, and es, and the three differences are computed by:
| $ \left\{ {dcu=(ec−eu)mod256dcd=(ec−ed)mod256dcs=(ec−es)mod256} \right. $ | (19) |
Step 2: After the differences of all the blocks have been computed, three small difference histograms are formed by using dcu, dcd, and dcs, respectively.
Step 3: Compute the embedding capacities of the three small difference histograms and choose the one that can provide the largest capacity for embedding data.
Step 4: After histogram shifting, an additional two bits should be used to denote which small difference histogram is used for embedding data. The additional two bits are embedded into two LSBs of one denoted pixel that does not belong to any T-type block, and the two LSBs are embedded together with the secret data to recover the image. The denoted pixel can be chosen pseudo-randomly by using the data hiding key from all the pixels that does not belong to any T-type block.
Step 5: The difference histogram shifting scheme of Li et al. [24], described in Section 2.2.2, is performed a limited number of rounds for embedding data. The limited number of rounds should be set in advance. To shift the elements in the difference histogram, the new value of the central pixel is computed to replace the original value of the central pixel. For example, if the difference histogram formed by dcu is used for embedding data, to shift the bin one step to the right, according to Eq. (12):
| $ {d_{{\text{cu}}}} + 1 = (({e_{\text{c}}} + 1)\bmod256 - {e_{\text{u}}})\bmod 256 $ | (20) |
Therefore, the central pixel, ec, is replaced with (ec+1) mod 256 for histogram shifting.
The computation of the capacity of the small difference histogram in Step 2 depends on the set of the limited round number for the difference histogram shifting. For example, if the limited round number is set to 3, in the original difference histogram, bins 0 and 255 will be used for data embedding in the first round, bins 1 and 254 will be used in the second round, and bins 2 and 253 will be used in the last round. Therefore, the capacity of the difference histogram is equal to the number of all the elements in bins 0–2 and bins 253–255.
Figure 5 shows an example of data embedding in the first round. First, three difference histograms are generated from the encrypted image, Lena (Figure 8(d)). As shown in the figure, in each histogram, most of the differences also are closed on the two sides, so the difference histogram shifting scheme of Li et al. [24] can be used for embedding data. Then, the difference histogram formed by dcu is chosen for embedding data, since it provides the largest capacity. Three rounds of difference histogram shifting are performed to embed the secret data.
In the second phase, the three neighboring pixels of the central pixel in each T-type block are used for embedding data. The process of embedding data in the second phase is proposed as follows:
Step 1: For each encrypted T-type block, denote the central pixel as $e_c^, $, and compute the three differences as:
| $ \left\{ {duc=(eu−e,c)mod256ddc=(ed−e,c)mod256dsc=(es−e,c)mod256} \right. $ | (21) |
Step 2: After the differences of all of the blocks have been computed, a large difference histogram is formed by using duc, ddc, and dsc together.
Step 3: Based on the large difference histogram, the data are embedded in the way that Li et al. [24] described in Section 2.2.2. To shift the bins one step in the difference histogram, eu, ed, and es are replaced with $({e_{\text{u}}} \pm 1)\bmod 256$, $({e_{\text{d}}} \pm 1)\bmod 256$, and $({e_{\text{s}}} \pm 1)\bmod 256$.
In the second phase, the first embedded bit is used to identify whether all the secret data have been embedded in the first phase. If there is no secret data to be embedded in the second phase, only a bit 0 is embedded in the second phase. Otherwise, a bit 1 is embedded first, then the rest secret data are embedded.
Figure 6 shows an example of embedding data in the second phase. The figure shows that, in the large difference histogram, most of the differences are still close to the two sides. Therefore, the difference histogram shifting of Li et al. [24] can be performed efficiently for embedding data.
After the marked encrypted image is transmitted, the receiver can extract the secret data and recover the image in the encrypted domain or the decrypted domain.
(1) Processing in the encrypted domain: Since the two-phase histogram shifting is performed directly on the encrypted image, the receiver can extract the embedded data and recover the encrypted image by performing the reverse processing of the two-phase histogram shifting as follows:
Step 1: Generate the large difference histogram according to Eq. (21).
Step 2: Based on the large difference histogram, extract the secret data that were embedded in the second phase of embedding data and recover the up pixel, eu, the down pixel, ed, and the side pixel, es, in each encrypted T-type block. The processes of data extraction and pixel recovery are the same as those of Li et al.'s method, as described in Section 2.2.3.
Step 3: Generate three small difference histogram according to Eq. (19) and then extract the additional two bits from the denoted pixel to determine the one that was used for embedding data.
Step 4: Based on the small difference histogram, extract the two LSBs and the secret data, shift the bins back to recover the central pixel ec in each T type block, and replace the two LSBs to recover the denoted pixel. Finally, all of the embedded data and the recovered encrypted image, which is identical to the original image, are retrieved.
Step 5: Decrypt the recovered encrypted image to obtain the original plain image.
(2) Processing in the decrypted domain: The receiver can decrypt the marked encrypted image, IME, directly to obtain the marked decrypted image, which is very similar to the original image.
Figure 7 shows a comparison of the large and small difference histograms generated from the marked encrypted image and the marked decrypted image. As shown in the figure, the two difference histograms generated from the decrypted image are identical to those generated from the encrypted image. In addition, according to the Eq. (14), the modification of the marked encrypted pixel is the same as the modification of the marked decrypted pixel due to additive homomorphism. Therefore, the processes of data extraction and image recovery performed on the marked decrypted image are the same as those performed on the marked encrypted image.
In this section, the proposed method is evaluated in terms of security, capacity, and visual quality. The two existing RDHEI methods, i.e., Li et al. [24] and Xiao et al. [25], are used for comparison. The six $512 \times 512$ standard grayscale test images used in the experiments are shown in Figure 8.
Figure 9 shows an example of the images in the process of the proposed method. Figure 9(b) shows the marked encrypted image with 0.5 bpp. The image appears as noise, so no meaningful content can be obtained. Figure 9(c) shows the decrypted version of Figure 9(b) with PSNR = 37.28 dB, and this version is similar to the original image. Figure 9(d) is the recovered image, which is identical to the original image.
Table 1 shows the PSNRs of the encrypted images and the correlation coefficients of the encrypted images with the plain images. The table shows that the PSNRs of the encrypted images are very low, and the correlation coefficients are almost zero. That means it was very hard to retrieve meaningful content from the encrypted image without the encryption key. For each $512 \times 512$ standard test image, the encryption key space of the encryption scheme in the proposed method is ${2^{255 \times 256 \times 256}}$. The encryption key space is very large so that the computational security can be achieved. Therefore the encryption scheme of the proposed method is secure.images and their correlation coefficients with the plain images.
| PSNR | correlation coefficients | |
| Airplane | 8.0554 | 0.0011 |
| Barbara | 8.8264 | 0.0000 |
| Lake | 8.3721 | 0.0044 |
| Lena | 9.2369 | 0.0026 |
| Peppers | 8.4663 | 0.0019 |
| Zelda | 8.8969 | -0.0021 |
DownLoad:
CSV
Table 2 compares the embedding capacities of the proposed method and the two existing RDHEI methods. As shown in the table, compared with the competitors, the proposed method can provide a higher embedding capacity. Compared with Li et al. [24], the T-type division and the two-phase histogram shifting of the proposed method fully exploit the pixels to generate more differences for embedding data. That clearly indicates that the proposed method improves the capacity.
| Xiao et al. [24] | Li et al. [25] | Proposed | |
| Airplane | 0.2764 | 0.7389 | 0.8763 |
| Barbara | 0.1670 | 0.6144 | 0.7236 |
| Lake | 0.1569 | 0.7379 | 0.7967 |
| Lena | 0.2398 | 0.7752 | 0.8950 |
| Peppers | 0.2136 | 0.7510 | 0.8479 |
| Zelda | 0.2670 | 0.7876 | 0.9206 |
DownLoad:
CSV
Figure 10 shows a comparison of the visual quality of the marked decrypted images with different embedding rates. As shown in the figure, the results of the proposed method is a little worse than that of Li et al.'s method [24] at the low embedding rates. However, as the embedding rate increased, the performance of the proposed method was very close to that of Li et al. [24]. For some images, such as Airplane and Barbara, the proposed method outperformed Li et al.'s [24] method at high embedding rates. In the proposed method, the elements of the bins at the two sides of the difference histogram are reduced, while the elements that are close to the two sides of the difference histogram are increased. When the embedding rate is low, Li et al.'s [24] technique requires performing fewer rounds of histogram shifting than the proposed method. However, when the embedding rate is high, the proposed method performs less rounds of histogram shifting than Li et al.'s scheme [24]. With the increase of the rounds of histogram shifting, the differences between the modified pixels and original pixels become larger. Performing more rounds of histogram shifting results in the greater distortion of the marked image. Therefore, as the embedding rate increases, the distortion of the image in the proposed method becomes close to that in Li et al. [24], and it may be even lower. In most cases, the performance of the proposed method was better than or similar to the performance of Xiao et al.'s method [25].
In this paper, a new RDHEI method based on two-phase histogram shifting was proposed. To fully exploit the pixels to reserve embedding capacity, a T-type division scheme and the RC4 additive homomorphic cryptosystem are used to generate the encrypted image which maintains the spatial correlation. From the perspective of the data hider, embedding data is performed in two phases. In each phase, based on the encrypted image, a difference histogram is generated, and the difference histogram shifting is used for embedding data. The extraction of data and the recovery of the image can be performed in both the decrypted and encrypted domains. The embedded data can be extracted without errors, and the recovered image is identical to the original image. The experimental results proved that the proposed method can provide high capacity for data embedding and maintain the visual quality of the marked decrypted image.
The main limitation of the proposed method is the relatively low visual quality of marked images at the low embedding rate. In the future work, to overcome the limitation and improve the visual quality, we aim to design a new dynamic strategy for the proposed method, which can efficiently distribute the secret data to the two phases of the data embedding process.
This paper is supported by the Natural Science Foundation of Fujian Province, China (No. 2017J05104, 2019H0021), the National Natural Science Foundation of China (No. 61701191), and the Xiamen Foundation for Science and Technology (No. 3502Z20173028).
The authors declare no conflicts of interest.
| [1] | Jones A (2000) Effects of cattle grazing on North American ecosystems: a quantitative review. West N Am Naturalist 60: 155-164. |
| [2] |
DiTomaso JM (2000) Invasive weeds in rangelands: Species, impacts, and management. Weed Sci 48: 255-265. doi: 10.1614/0043-1745(2000)048[0255:IWIRSI]2.0.CO;2
|
| [3] | Miller RF, Bates JD, Svejcar TJ, et al. (2005) Biology, ecology, and management of western juniper. Tech Bull 152. Corvallis, OR: Oregon State University, Agricultural Experiment Station. 82 p. |
| [4] | Miller RF, Eddleman LE (2001) Spatial and temporal changes of sage grouse habitat in the sagebrush biome. Tech Bull 151. Corvallis, OR: Oregon State University Agricultural Experiment Station. |
| [5] |
Schroeder MA, Aldridge CL, Apa AD, et al. (2004) Distribution of sage-grouse in North America. Condor 106: 363-376. doi: 10.1650/7425
|
| [6] | Connelly JW, Knick ST, Schroeder MA, et al. (2004) Conservation assessment of Greater sage-grouse and sagebrush habitats. Cheyenne, WY: Western Association of Fish and Wildlife Agencies. 610 p. |
| [7] | Braun CE, Connelly JW, Schroeder MA (2005) Seasonal habitat requirements for sage-grouse: spring, summer, fall, and winter. In: Shaw NL, Pellant M, Monsen SB, editors. Sage-grouse habitat restoration symposium proceedings: June 4-7, Boise, ID Proc RMRS-P-38. Fort Collins, CO: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station Proceedings RMRS-P-38. pp. 130. |
| [8] | Knick ST, Connelly JW, editors (2011) Greater sage-grouse: ecology and conservation of a landscape species and its habitats. Studies in Avian Biology 38. Berkeley, CA: University of California Press. |
| [9] | Crawford JA, Olson RA, West NE, et al. (2004) Ecology and management of sage-grouse and sage-grouse habitat. J Range Manage 57: 2-19. |
| [10] | US Department of the Interior (2010) Endangered and threatened wildlife and plants; 12-month findings for petitions to list the greater sage-grouse (Centrocercus urophasianus) as threatened or endangered. Federal Register 75:13910-14014 (23 March 2010). |
| [11] | Mack RN (1981) Invasion of Bromus tectorum L. into western North America: an ecological chronicle. Agro-Ecosystems 7: 145-165. |
| [12] |
Soulé PT, Knapp PA (1999) Western juniper expansion on adjacent disturbed and near-relict sites. J Range Manage 52: 525-533. doi: 10.2307/4003782
|
| [13] |
Miller RF, Svejcar TJ, Rose JA (2000) Impacts of western juniper on plant community composition and structure. J Range Manage 53: 574-585. doi: 10.2307/4003150
|
| [14] | Pellant M (1996) Cheatgrass: the invader that won the West. Bureau of Land Management, Idaho State Office. 23 p. |
| [15] | Whisenant SG (1990) Changing fire frequencies on Idaho's Snake River Plains: ecological and management implications. In: McArthur ED, Romney EM, Smith SD et al., editors. Proceedings: symposium on cheatgrass invasion, shrub die-off, and other aspects of shrub biology and management. Gen Tech Rep INT-276. Gen Tech Rep INT-276 ed. Ogden, UT: U.S. Department of Agriculture, Forest Service, Intermountain Research Station. pp. 4-10. |
| [16] | Miller RF, Rose JA (1995) Historic expansion of Juniperus occidentalis (western juniper) in southeastern Oregon. Great Basin Nat 55: 37-45. |
| [17] |
Burkhardt JW, Tisdale EW (1976) Causes of juniper invasion in southwestern Idaho. Ecology 57: 472-484. doi: 10.2307/1936432
|
| [18] | Bunting SC, Kingery JL, Strand E (1999) Effects of succession on species richness of the western juniper woodland/sagebrush steppe mosaic. In: Monsen SB, Richards S, Tausch RJ et al., editors. Proceedings: Ecology and management of piñon-juniper communities within the Interior West: USDA Forest Service, Rocky Mountain Research Station-P-9. |
| [19] | Mote P, Snover AK, Capalbo S, et al. (2014) Ch. 21: Northwest. In: Melillo JM, Richmond TC, Yohe GW, editors. Climate Change Impacts in the United States: The Third National Climate Assessment: U.S. Global Change Research Program. |
| [20] |
Abatzoglou JT, Kolden CA (2011) Climate change in western US deserts: potential for increased wildfire and invasive annual grasses. Rangeland Ecol Manag 64: 471-478. doi: 10.2111/REM-D-09-00151.1
|
| [21] |
Abatzoglou JT, Rupp DE, Mote PW (2014) Seasonal climate variability and change in the Pacific Northwest of the United States. J Climate 27: 2125-2142. doi: 10.1175/JCLI-D-13-00218.1
|
| [22] |
Bradley BA, Oppenheimer M, Wilcove DS (2009) Climate change and plant invasions: restoration opportunities ahead? Global Change Biol 15: 1511-1521. doi: 10.1111/j.1365-2486.2008.01824.x
|
| [23] |
Beschta RL, Donahue DL, DellaSala DA, et al. (2013) Adapting to climate change on western public lands: addressing the ecological effects of domestic, wild, and feral ungulates. Environ Manage 51: 474-491. doi: 10.1007/s00267-012-9964-9
|
| [24] | Chambers JC, Bradley BA, Brown CS, et al. (2014) Resilience to stress and disturbance, and resistance to Bromus tectorum L. invasion in cold desert shrublands of western North America. Ecosystems 17: 360-375. |
| [25] | McIver J, Starr L (2001) Restoration of degraded lands in the interior Columbia River basin: passive vs. active approaches. For Ecol Manage 153: 15-28. |
| [26] | Creutzburg MK, Halofsky JE, Halofsky JS, et al. (2014) Climate change and land management in the rangelands of central Oregon. Environ Manage 55: 43-55. |
| [27] |
Halofsky JS, Halofsky JE, Burcsu T, et al. (2014) Dry forest resilience varies under simulated climate-management scenarios in a central Oregon, USA landscape. Ecol Appl 24:1908-1925. doi: 10.1890/13-1653.1
|
| [28] | Yospin GI, Bridgham SD, Neilson RP, et al. (2014) A new model to simulate climate change impacts on forest succession for local land management. Ecol Appl 25: 226-242. |
| [29] | Bachelet D, Lenihan JM, Daly C, et al. (2001) MC1, a dynamic vegetation model for estimating the distribution of vegetation and associated carbon and nutrient fluxes, Technical Documentation Version 1.0. Portland, Oregon, USA: USDA Forest Service, Pacific Northwest Station. |
| [30] | Neilson RP (1995) A model for predicting continental-scale vegetation distribution and water balance. Ecol Appl 5: 352-385. |
| [31] |
Parton WJ, Scurlock JMO, Ojima DS, et al. (1993) Observations and modeling of biomass and soil organic-matter dynamics for the grassland biome worldwide. Global Biogeochem Cy 7: 785-809. doi: 10.1029/93GB02042
|
| [32] | Lenihan JM, Daly C, Bachelet D, et al. (1998) Simulating broad-scale fire severity in a dynamic global vegetation model. Northwest Sci 72: 91-103. |
| [33] |
Halofsky JE, Hemstrom MA, Conklin DR, et al. (2013) Assessing potential climate change effects on vegetation using a linked model approach. Ecol Model 266: 131-143. doi: 10.1016/j.ecolmodel.2013.07.003
|
| [34] | Breiman L, Priedman J, Stone CJ, et al. (1984) Classification and Regression Trees. Boca Raton, FL: Chapman & Hall. 359 p. |
| [35] |
Bestelmeyer BT, Tugel AJ, Peacock Jr. GL, et al. (2009) State-and-transition models for heterogeneous landscapes: a strategy for development and application. Rangeland Ecol Manag 62: 1-15. doi: 10.2111/08-146
|
| [36] |
Westoby M, Walker B, Noy-Mier I (1989) Opportunistic management for rangelands not at equilibrium. J Range Manage 42: 66-274. doi: 10.2307/3899661
|
| [37] |
Briske DD, Fuhlendorf SD, Smeins FE (2005) State-and-transition models, thresholds, and rangeland health: a synthesis of ecological concepts and perspectives. Rangeland Ecol Manag 58: 1-10. doi: 10.2111/1551-5028(2005)58<1:SMTARH>2.0.CO;2
|
| [38] | Briske DD, Bestelmeyer BT, Stringham TK, et al. (2006) Recommendations for development of resilience-based state-and-transition models. Rangeland Ecol Manag 61: 359-367. |
| [39] |
Provencher L, Forbis TA, Frid L, et al. (2007) Comparing alternative management strategies of fire, grazing, and weed control using spatial modeling. Ecol Model 209: 249-263. doi: 10.1016/j.ecolmodel.2007.06.030
|
| [40] |
Forbis TA, Provencher L, Frid L, et al. (2006) Great Basin land management planning using ecological modeling. Environ Manage 38: 62-83. doi: 10.1007/s00267-005-0089-2
|
| [41] |
Hemstrom MA, Merzenich J, Reger A, et al. (2007) Integrated analysis of landscape management scenarios using state and transition models in the upper Grande Ronde River subbasin, Oregon, USA. Landscape Urban Plan 80: 198-211. doi: 10.1016/j.landurbplan.2006.10.004
|
| [42] |
Hemstrom MA, Wisdom MJ, Hann WJ, et al. (2002) Sagebrush-steppe vegetation dynamics and restoration potential in the interior Columbia Basin, U.S.A. Con Bio 16: 1243-1255. doi: 10.1046/j.1523-1739.2002.01075.x
|
| [43] | Creutzburg MK, Halofsky JS, Hemstrom MA (2012) Using state-and-transition models to project cheatgrass and juniper invasion in eastern Oregon sagebrush steppe. In: Kerns BK, Shlisky A, Daniel CJ, editors. Proceedings of the First State-and-Transition Landscape Modeling Conference, June 15-16, 2011. Gen Tech Rep PNW-869. Portland, OR: U.S. Department of Agriculture, Forest Service, Pacific Northwest Forest and Range Experiment Station. |
| [44] |
Littell JS, Oneil EE, McKenzie D, et al. (2010) Forest ecosystems, disturbance, and climatic change in Washington State, USA. Climatic Change 102: 129-158. doi: 10.1007/s10584-010-9858-x
|
| [45] | U.S. Geological Survey and U.S. Department of Agriculture NRCS (2011) Federal Standards and Procedures for the National Watershed Boundary Dataset (WBD) (2d ed.). U.S. Geological Survey Techniques and Methods. pp. 62 p. |
| [46] | Burcsu TK, Halofsky JS, Bisrat SA, et al. (2014) Chapter 2: Dynamic Vegetation Modeling of Forested, Woodland, Shrubland, and Grassland Vegetation Communities in the Pacific Northwest and Southwest Regions of the United States. In: Halofsky JE, Creutzburg MK, Hemstrom MA, editors. Integrating Social, Economic, and Ecological Values across Large Landscapes. Gen Tech Rep PNW-GTR-896. Portland, OR: U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station. 216 pp.. |
| [47] |
Taylor KE, Stouffer RJ, Meehl GA (2012) An overview of CMIP5 and the experiment design. B Am Meteorol Soc 93: 485-498. doi: 10.1175/BAMS-D-11-00094.1
|
| [48] | Connelly JW, Schroeder MA, Sands AR, et al. (2000) Guidelines to manage sage grouse populations and their habitats. Wildlife Soc B 28: 967-985. |
| [49] |
Anderson JE, Inouye RS (2001) Landscape-scale changes in plant species abundance and biodiversity of a sagebrush steppe over 45 years. Ecol Monogr 71: 531-556. doi: 10.1890/0012-9615(2001)071[0531:LSCIPS]2.0.CO;2
|
| [50] |
Cook JG, Irwin LL (1992) Climate-Vegetation Relationships between the Great-Plains and Great-Basin. Am Midl Nat 127: 316-326. doi: 10.2307/2426538
|
| [51] |
Nagler PL, Glenn EP, Kim H, et al. (2007) Relationship between evapotranspiration and precipitation pulses in a semiarid rangeland estimated by moisture flux towers and MODIS vegetation indices. J Arid Environ 70: 443-462. doi: 10.1016/j.jaridenv.2006.12.026
|
| [52] |
Bates JD, Svejcar T, Miller RF, et al. (2006) The effects of precipitation timing on sagebrush steppe vegetation. J Arid Environ 64: 670-697. doi: 10.1016/j.jaridenv.2005.06.026
|
| [53] |
Bradley BA (2009) Regional analysis of the impacts of climate change on cheatgrass invasion shows potential risk and opportunity. Global Change Biol 15: 196-208. doi: 10.1111/j.1365-2486.2008.01709.x
|
| [54] |
Chambers JC, Roundy BA, Blank RR, et al. (2007) What makes Great Basin sagebrush ecosystems invasible by Bromus tectorum? Ecol Monogr 77: 117-145. doi: 10.1890/05-1991
|
| [55] | Parton WJ, Anderson DW, Cole CV, et al. (1983) Simulation of soil organic matter formation and mineralization in semiarid agroecosystems. Special Publication No. 23. In: Lowrance RR, Todd RL, Asmussen LE et al., editors. Nutrient cycling in agricultural ecosystems. Athens, Georgia: The University of Georgia, College of Agriculture Experiment Stations. |
| [56] | Rogers BM, Neilson RP, Drapek R, et al. (2011) Impacts of climate change on fire regimes and carbon stocks of the U.S. Pacific Northwest. J Geophys Res-Biogeo 116: G03037. |
| [57] | Michalak JL, Withey JC, Lawler JJ, et al. (2014) Climate vulnerability and adaptation in the Columbia Plateau, WA. Report prepared for the Great Northern Landscape Conservation Cooperative. Available from:https://www.sciencebase.gov/catalog/item/533c5408e4b0f4f326e3a15e. |
| [58] |
Aldridge CL, Nielsen SE, Beyer HL, et al. (2008) Range-wide patterns of greater sage-grouse persistence. Divers Distrib 14: 983-994. doi: 10.1111/j.1472-4642.2008.00502.x
|
| [59] | Innes RJ (2013) Fire Effects Information System, [Online]. U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station, Fire Sciences Laboratory (Producer). Available from: http://www.fs.fed.us/database/feis. |
| [60] |
McKenzie D, Gedalof Z, Peterson DL, et al. (2004) Climatic change, wildfire, and conservation. Conserv Biol 18: 890-902. doi: 10.1111/j.1523-1739.2004.00492.x
|
| [61] |
Brooks ML, D'Antonio CM, Richardson DM, et al. (2004) Effects of invasive alien plants on fire regimes. BioScience 54: 677-688. doi: 10.1641/0006-3568(2004)054[0677:EOIAPO]2.0.CO;2
|
| [62] | Connelly JW, Arthur WJ, Markham OD (1981) Sage grouse leks on recently disturbed sites. 34: 153-154. |
| [63] | Davies G, Bakker J, Dettweiler-Robinson E, et al. (2012) Trajectories of change in sagebrush steppe vegetation communities in relation to multiple wildfires. Ecol Appl 22: 1562-1577. |
| [64] | Holloran MJ, Anderson SH (2005) Spatial distribution of greater sage-grouse nests in relatively contiguous sagebrush habitats. Condor 107: 742-752. |
| [65] |
Kremer RG, Hunt ER, Running SW, et al. (1996) Simulating vegetational and hydrologic responses to natural climatic variation and GCM-predicted climate change in a semi-arid ecosystem in Washington, USA. J Arid Environ 33: 23-38. doi: 10.1006/jare.1996.0043
|
| [66] |
Ziska LH, Reeves JB, Blank B (2005) The impact of recent increases in atmospheric CO2 on biomass production and vegetative retention of Cheatgrass (Bromus tectorum): implications for fire disturbance. Global Change Biol 11: 1325-1332. doi: 10.1111/j.1365-2486.2005.00992.x
|
| [67] |
Smith SD, Huxman TE, Zitzer SF, et al. (2000) Elevated CO2 increases productivity and invasive species success in an arid ecosystem. Nature 408: 79-82. doi: 10.1038/35040544
|
| [68] |
Bradley BA, Wilcove DS (2009) When invasive plants disappear: transformative restoration possibilities in the western United States resulting from climate change. Restor Ecol 17: 715-721. doi: 10.1111/j.1526-100X.2009.00586.x
|
| [69] |
Baruch-Mordo S, Evans JS, Severson JP, et al. (2013) Saving sage-grouse from the trees: A proactive solution to reducing a key threat to a candidate species. Biol Conserv 167: 233-241. doi: 10.1016/j.biocon.2013.08.017
|
| [70] | Boyd CS, Johnson DD, Kerby J, et al. (2014) Of grouse and golden eggs: can ecosystems be managed within a species-based regulatory framework? Range Ecol Manag 57: 358-368. |
| [71] |
Davies KW, Sheley RL (2011) Promoting native vegetation and diversity in exotic annual grass infestations. Restor Ecol 19: 159-165. doi: 10.1111/j.1526-100X.2009.00548.x
|
| [72] |
Pyke DA, Wirth TA, Beyers JL (2013) Does seeding after wildfires in rangelands reduce erosion or invasive species? Restor Ecol 21: 415–421. doi: 10.1111/rec.12021
|
| [73] |
Davies KW, Johnson DA (2011) Are we “missing the boat” on preventing the spread of invasive plants in rangelands? Invasive Plant Sci Manag 4: 166-171. doi: 10.1614/IPSM-D-10-00030.1
|
| [74] |
Drut MS, Pyle WH, Crawford JA (1994) Technical note: Diets and food selection of sage grouse chicks in Oregon. J Range Manage 47: 90-93. doi: 10.2307/4002848
|
| [75] | Knutti R, Sedlacek J (2013) Robustness and uncertainties in the new CMIP5 climate model projections. Nat Clim Change 3: 369-373. |
| [76] | Huntley B (1991) How plants respond to climate change - migration rates, individualism and the consequences for plant-communities. Ann Bot-London 67: 15-22. |
| 1. | Soo-Mok Jung, Byung-Won On, An Advanced Reversible Data Hiding Algorithm Using Local Similarity, Curved Surface Characteristics, and Edge Characteristics in Images, 2020, 10, 2076-3417, 836, 10.3390/app10030836 | |
| 2. | Yongsheng Ding, Yunbo Wei, Shuisheng Zhang, Shihang Yu, Yuvaraja Teekaraman, Reversible Information of Encrypted Image Based on Feature Difference Detection and Wavelet Transform, 2021, 2021, 1555-4317, 1, 10.1155/2021/5483001 | |
| 3. | Guangyong Gao, Hui Zhang, Zhihua Xia, Xiangyang Luo, Yun-Qing Shi, Reversible data hiding-based contrast enhancement with multi-group stretching for ROI of medical image, 2023, 1520-9210, 1, 10.1109/TMM.2023.3318048 | |
| 4. | Guangyong Gao, Sitian Yang, Xiangyang Hu, Zhihua Xia, Yun-Qing Shi, Reversible Data Hiding-Based Local Contrast Enhancement With Nonuniform Superpixel Blocks for Medical Images, 2025, 35, 1051-8215, 1745, 10.1109/TCSVT.2024.3482556 |
Figures(8) / Tables(4)
Megan K. Creutzburg, Emilie B. Henderson, David R. Conklin. Climate change and land management impact rangeland condition and sage-grouse habitat in southeastern Oregon[J]. AIMS Environmental Science, 2015, 2(2): 203-236. doi: 10.3934/environsci.2015.2.203
| PSNR | correlation coefficients | |
| Airplane | 8.0554 | 0.0011 |
| Barbara | 8.8264 | 0.0000 |
| Lake | 8.3721 | 0.0044 |
| Lena | 9.2369 | 0.0026 |
| Peppers | 8.4663 | 0.0019 |
| Zelda | 8.8969 | -0.0021 |
DownLoad:
CSV
| PSNR | correlation coefficients | |
| Airplane | 8.0554 | 0.0011 |
| Barbara | 8.8264 | 0.0000 |
| Lake | 8.3721 | 0.0044 |
| Lena | 9.2369 | 0.0026 |
| Peppers | 8.4663 | 0.0019 |
| Zelda | 8.8969 | -0.0021 |
| Xiao et al. [24] | Li et al. [25] | Proposed | |
| Airplane | 0.2764 | 0.7389 | 0.8763 |
| Barbara | 0.1670 | 0.6144 | 0.7236 |
| Lake | 0.1569 | 0.7379 | 0.7967 |
| Lena | 0.2398 | 0.7752 | 0.8950 |
| Peppers | 0.2136 | 0.7510 | 0.8479 |
| Zelda | 0.2670 | 0.7876 | 0.9206 |
