Techniques

Author & Reference

Year

Description

Quality measurement

Advantages/Disadvantages

Preprocessing Techniques

2-nd level Daubechies (db4) Discrete Wavelet Transform

Carvajal-Gamez, et al [13]

2013

The pixels are classified using threshold based on the standard deviation of the local complexity of the cover image. Three types of color spaces are used to provide the difference between cover images and stego images

1. Peak Signal to Noise Ratio (PSNR)

2. Mean Absolute Error (MAE)

3. Normalized Color Difference (NCD)

4. Cross-Correlation (CC)

5. Quality index (Q)

6. Hiding Capacity (HC)

Advantage

1. Efficient and feasible for adaptive steganography applications

Disadvantage

1. Less probability to detect hidden data

Vector Rank M-type L (VRML)-filter

Gallegos-Funes, et al. [14]

2012

The impulsive noise detector is presented to enhance the properties of noise suppression. It combines L algorithm with the Rank M-type(RM) estimator

1. Peak Signal to Noise Ratio (PSNR)

2. Mean Absolute Error (MAE)

3. Mean Chromaticity Error (MCRE)

4. Normalized Color Difference (NCD)

Advantage

1. Better impulsive noise suppression

Disadvantage

1. Increased computational complexity

Successive Mean Quantization Transform (SMQT)

Sajedi and Jamzad [15]

2010

The cover image is processed into two stages such as preprocessing and embedding. An automatic structural breakdown of information is achieved. The organization of the structure the data is revealed by removing gain and bias properties

1. Detection accuracy

2. Embedding capacity

Advantage

1. Enhanced image quality

2. Improved security

Disadvantage

1. Detection is not reliable

2. Local enhancement is not detected easily

Quaternion Transform and Scale Invariant Feature Transform (SIFT)

Hussain, et al. [16]

2016

The textural features are extracted in both online and offline mode. In online, the large data sets are obtained from website and search engines. In offline, data sets are acquired from large stored database

1. Region Of Interest (ROI)

2. Reduction in number of regions

3. Reduction in SIFT features

Advantage

1. Improved accuracy

2. Increased efficiency

Disadvantage

1. Reduced speed

2. The complex images are not considered

Anisotropic Diffusion (AD) technique

Abhishree, et al. [17]

2015

The successive diffusion is performed on the basis of more and more blurring image. the entire process is divided into preprocessing, feature extraction and feature selection

1. Variation in Pose

2. Illumination

3. Face expression

Advantage

1. Reduced testing time

2. High recognition rate

Disadvantage

1. Not suitable for real time application

2. Low speed computational methods are used

Chirp Z-Transform (CZT) and Goertzel algorithm

Varadarajan, et al. [18]

2015

The input image is enhanced by applying CZT. The problem is optimized by improving the candidate solution with respect to their quality.

1. Average recognition rate

2. Number of selected features

3. testing time

Advantage

1. Optimal recognition rate

Disadvantage

1. Image enhancement is not satisfied

Feature Extraction Techniques

Logarithmic non-subsampled contourlet transform

(LNSCT)

Farhat and Ghaemmaghami [20]

2015

The strong edges, weak edges and noise are extracted from the images. Every images are distributed into five subsets regarding the angle between light source direction and camera axis

1. Recognition rate

2. Equal Error Rate (EER)

Advantage

1. Robust extraction of features

Disadvantage

1. Slower computation time

2. Increased computational complexity

Discrete Cosine Transform (DCT)

Karimi, et al. [21]

2015

The original features are extended as the extended version to extract the more complicated relation among the coefficients. The ANOVA F-test is applied to compare the mean of sampled data.

1. Median testing error

2. Embedding rate

Advantage

1. Superior performance

2. The hidden data can be extracted with low embedding rate

Disadvantage

1. Increased dimensionality