Page 177 - 2024-Vol20-Issue2

173 |
Fig. 16. Results of the decrypted subscheme. (a) unscrambled
image, (b) de-fusion image, (c) decrypted Lena, (d) decrypted
baboon images.
TABLE I.
ENTROPY FOR VARIOUS COLOR IMAGES IN THE THREE
COLOR CHANNELS USED IN THE HDOE SCHEME.
Entropy
RGB Lena- Barbara- Sailboat-
Channels Peppers
Red Baboon Airplane 7.9982
Green 7.9975
Blue 7.9987 7.9978 7.9961
7.9984 7.9971
7.9986 7.9976
in Fig. 10, as shown in Fig. 18.
C. Correlation Coefficient Analysis
A correlation coefficient measures the similarity or dissimi-
larity between adjacent image pixels along the vertical, hori-
zontal, and diagonal directions. The correlation coefficient’s
value falls between -1 and 1, with -1 denoting a negative
correlation, +1 a positive correlation, and 0 denoting no corre-
lation. In order to withstand statistical attacks, the encrypted
image must have a correlation coefficient between neighbor-
ing pixels nearly equal to zero in all directions. The results are visual information. This indicates that the proposed method
depicted in Figs. 19 and 20, which present the three-channel
correlation coefficients for the original images of Lena and
Baboon, respectively. As can be observed, the adjacent pixels
in the three directions (horizontal (HD), vertical (VD), and
diagonal (DD)) have a linear correlation characteristic. This
is in contrast to the correlation characteristics of the encrypted
images Lena-baboon as displayed in Fig. 21. Table II lists
the three-direction correlation coefficient for the used origi-
nal images. The encrypted image’s horizontal, vertical, and
diagonal correlation coefficients are also given in Table III. In-
vestigation of these results reveals that the encrypted images’
Muttashar & Fyath
Fig. 17. Histograms of the RGB encrypted images.
correlation coefficients are almost zero. This highlights the
effectiveness of the encryption process adopted in this work.
D. Noise Attack Analysis
During transmission, noise significantly impacts the quality
of the encrypted image. The results show that the noise at-
tack with various noise levels on the encrypted image yields
decrypted images that still maintain noise resistance and im-
age retrieval well. This is illustrated in Fig. 22 where the
decrypted images of Lena, Baboon, sailboat, and peppers
are displayed when the corresponding encrypted images are
affected by a Gaussian noise of standard deviation s . The
results are presented for s = 0.5, 5, 10, 15, and 20 in columns
a-e, respectively. The results show that the proposed scheme
resists noise attacks.
E. Cropping Attack Analysis
To test the effectiveness of the proposed method against crop-
ping attacks, one can examine the performance of the en-
crypted image after the cropped with various formats, as
shown in Fig. 23. Parts a-d of this figure show the encrypted
image cropped with size 32×32 from the upper left, with size
64×64, with size 128×128 from the upper left, and with size
128×128 from the middle image right, respectively. The de-
crypted versions of the corresponding images are displayed in
parts (e)-(l) of the figure and show that the deciphered images
are still recognizable and keep the majority of the original
of encryption is resistant to occlusion attacks.