Table 1 Various loss functions and their purpose.

From: Modified energy-based GAN for intensity in homogeneity correction in brain MR images

S.No

Loss term

Purpose

Effect on output

1

Guidance L1

Pixel accuracy

Structural correctness

2

Adversarial

Visual realism

Sharp, realistic textures

3

Histogram correlation

Contrast/brightness distribution

Natural-looking lighting

4

3D Loss

Spatial coherence in volumes

3D or temporal consistency

5

Pulling-Away Term

Sample diversity

Prevents mode collapse