Table 1 Summary of Statistical Analysis of Textural Features

From: Detection of AI generated images using combined uncertainty measures and particle swarm optimised rejection mechanism

(a) Univariate Analysis (Welch’s t-statistic)

Dataset

Contrast

Energy

Entropy

Homogeneity

VQDM

24.50

-2.82

20.02

-19.57

Midjourney

-19.19

-8.31

0.32*

-5.40

BigGAN

26.20

-6.24

21.95

-20.19

StableDiffusion

-9.58

6.60

-7.91

5.07

Glide

-9.58

6.60

-7.91

5.07

(b) Multivariate Analysis (Hotelling’s T2)

Dataset

Hotelling’s T2

VQDM

795.16

Midjourney

661.19

BigGAN

849.95

StableDiffusion

155.04

Glide

155.04

(c) Distributional Drift (KL Divergence)

Dataset

Contrast

Energy

Entropy

Homogeneity

VQDM

0.01

39.57

2.54

36.94

Midjourney

0.00

12.68

0.83

7.48

BigGAN

0.02

36.69

6.95

33.67

StableDiffusion

0.00

58.15

1.22

43.47

Glide

0.00

58.15

1.22

43.47

  1. * \(p = 0.7477\), indicating no significant difference. For all other t-tests and all Hotelling’s T2 tests, \(p < 0.005\)