AdjustContrast does what commonly called gamma correction

[function2-llx]

Hello everyone,

Seems like what AdjustContrast transformation does is commonly called "gamma correction". The operation of contrast adjustment is commonly defined as multiplying some factor (possibly blending with (1-factor) * mean). The link provided in #114, torchvision implementation for adjust_contrast, adjust_gamma, and Pillow suggest the same naming convention.

This can be confusing sometimes, e.g., in #5535, when talking about AdjustContrast in MONAI, one may refer to adjust_contrast in torchvision (like this comment), while they do different stuffs.

Shall we consider renaming this? Or maybe we're supposed to used ColorJitter from torchvision in this case.

MONAI/monai/transforms/intensity/array.py

Lines 801 to 809 in f5708ea

	class AdjustContrast(Transform):
	"""
	Changes image intensity by gamma. Each pixel/voxel intensity is updated as::
	x = ((x - min) / intensity_range) ^ gamma * intensity_range + min
	Args:
	gamma: gamma value to adjust the contrast as function.
	"""

[KumoLiu]

Hi @function2-llx, we could already use PyTorch TorchVision transform based on the specified transform name and args. Here is an example. https://github.com/Project-MONAI/tutorials/blob/4d4c565c681fd7df0b01889c9d599b9b73e0b401/pathology/hovernet/training.py#L209-L216

For the naming convention, do you have some comments here @wyli?

[function2-llx]

@KumoLiu Thank you for the reply. Yes, we can use torchvision-based transform, though we cannot specify the probability (no prob parameter in RandTorchVisionD. Of course, I can implement a simple wrapper when necessary, or combine with RandLambda), and it also supports 2D only.