IEEE Consumer Electronics Magazine - May 2018 - 105

image that is taken as input. As 2-DDCT operates on 8 # 8 blocks, the
complete N # N matrix is divided into
multiple 8 # 8 submatrices or blocks.
Each element of the block represents
the pixel intensity level of the corresponding element of the block. The
pixel intensity of an eight-bit grayscale
image lies between 0 and 255, where
the minimum value represents pure
black and the maximum value represents pure white. JPEG image compression can be achieved by transforming
each input image block M through
2-D-DCT coefficient matrix T using
the following equation:
X = T ) M ) T trans,

(1)

where X is the corresponding discrete
cosine transformed block of input
image block M, and T trans is the transpose of 2-D-DCT coefficient matrix T.
Quantization is applied on the transformed block to generate a compressed
image block (X l ) . The compressed
image block generated in an 8 # 8 2-D
matrix form is converted into a one-

dimensional (1-D) array through zigzag
scanning [3]. The run-length encoding
is applied to generate the bit stream data
of the compressed image to store it in
a device. The complete process flow of
a JPEG compression technique is shown
in Figure 1.

PROPOSED OBFUSCATED
HARDWARE IP DESIGN FOR JPEG
IMAGE DECOMPRESSION
PROBLEM FORMULATION
For this project, the problem is for an
N # N grayscale input image, design an
obfuscated IP core hardware for JPEG
image decompression that is resilient
against reverse engineering attacks.

PROPOSED OBFUSCATED IP DESIGN
PROCESS FOR JPEG IMAGE
DECOMPRESSION
In the JPEG image decompression process, the bit stream of the compressed
image data stored in a device is decoded
through run-length decoding. The decoded 1-D array is used to reconstruct

the compressed 2-D image block (X l )
through inverse zigzag scanning. Inverse
quantization is applied on the compressed image block (X l ) to generate a
dequantized image block (X m ). Finally,
the decompressed image (O) can be
reconstructed by applying inverse transformation on each dequantized image
block X m through 2-D-DCT coefficient
matrix T using the following equation:
O = T trans ) X m ) T,

(2)

where O is the corresponding decompressed image block of dequantized
image block X m, and T trans is the transpose of 2-D-DCT coefficient matrix T.
To convert the matrix relationship into a
hardware function for dedicated IP core
design, the pixels (X m ij) of the dequantized image block (X m ) are transformed
into decompressed image pixels (o ij)
[using (2)], e.g., for o 11, which is the first
pixel of the decompressed image, such
as in (3).
The same structure of the aforementioned equation is yielded to calculate
other pixels of the decompressed image

Input Grayscale Image

Image to Matrix Conversion

Levelizing the Data (-128)

Transformation Using 2-D-DCT Coefficient Matrix

Perform Quantization on Transformed Matrix
JPEG Compression
Zigzag Scanning of Quantized Matrix

Perform Run-Length Encoding
Store Compressed
Image Data

FIGURE 1. The complete process flow of a JPEG compression technique.

may 2018

IEEE Consumer Electronics Magazine

105

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - May 2018