![](https://csdnimg.cn/release/download_crawler_static/83236539/bg3.jpg)
while the second proposed JFD method uses a dynamic
key-trees approach to replace permutation, which is more
flexible but requires more information (a session key).
The operations of decryption and fingerprinting on the
receiver side recover the encrypted image according to
the set of pre-designed key-trees. The set of elements of
key-trees is unique (i.e., codebook) and can save signifi-
cant bandwidth and conveniently update key-trees. The
experimental results show that the encrypted images are
unintelligible in visual perception and the fingerprinted
images possess desirable visual quality.
The rest of this paper is organized as follows. In Section
2, techniques used in this paper will be introduced.
Section 3 details the proposed JFD scheme based on vector
quantization with two encryption methods. Then, the
experimental results will be given in Section 4. Finally,
conclusions are drawn in Section 5.
2. Review of techniques used
2.1. Vector quantization
Vector quantization (VQ) [18,19] is a classical lossy
compression technique with simple and efficient encod-
ing and decoding procedures, which is especially designed
for digital images. Fig. 3 shows the flow chart of the VQ
encoding and decoding process.
In the encoding process, the original image is first
divided into several non-overlapping blocks of w h
pixels. That is, each block can be denoted by a w h-
dimensional vector. The main concept of VQ is to map
each image block using a mapping function G from w h-
dimensional Euclidean space R
w h
to a finite subset CB of
R
w h
, i.e., G: R
w h
-CB, where CB¼ {Y
1
, Y
2
,y, Y
L
} is called
the codebook and Y
i
is the i-th codeword in CB and L refers
to the size of the codebook. Each codeword Y
i
corresponds
to w h elements; that is, Y
i
¼{y
1
, y
2
,y, y
w h
}.
For each vector XAR
w h
of the original image, the
nearest codeword in the codebook can be found according
to the distance between X and a codeword Y
i
, i¼1, 2, y, L.
The distance is determined by the following Euclidean
distance, d(X, Y
i
):
dðX, Y
i
Þ¼:XY
i
: ¼
X
wh
j ¼ 1
ðx
j
y
i, j
Þ
2
2
4
3
5
1=2
, ð1Þ
where x
j
and y
i
,
j
represents the j-th elements of vectors X
and Y
i
, respectively. When the nearest codeword Y
i
of X is
found, index i is used to encode vector X. After all X’s are
encoded, the original image can be eventually represented
by indices of these nearest codewords.
In the decoding process, the image is reconstructed by
table lookup operations. The same codebook that is used
by the VQ encoder is also required by the VQ decoder.
According to the indices represented by the VQ encoder,
the VQ decoder uses the corresponding codewords to
reconstruct the image. Therefore, the quality of the VQ-
compressed image is significantly influenced by the quality
of the codebook, which can be well-designed by [19,20].
2.2. The strength secure session key for media distribution
In a symmetric encryption scenario as shown in Fig. 4,
the sender and receiver must share the same key for
communication or transmission, and that key must be
protected from access by others. A permanent key, which
may be used without alterations for years, is convenient
for users but is easily exposed to attackers, who can learn
the key. Session keys can solve this problem.
Unlike typical keys or master keys, a session key is
created to exist for a limited period of time. As shown in
Fig. 5, the sender encrypts the original data with the
session key and then encrypts the session key with the
Nearest
Codeword
Index Matrix Table Lookup
xxxxxxxx
xxxxxxxx
xxxxxxxx
xxxxxxxx
Codebook
Index i
Original Block Recovered Block
h
w w
h
Original Image Recovered Image
Encoding Decoding
Fig. 3. VQ encoding and decoding.
C.-Y. Lin et al. / Signal Processing 92 (2012) 2159–2171 2161