Source-linked AI summary
A Comprehensive Review of Image Enhancement Techniques
Raman Maini, Himanshu Aggarwal
TL;DR
Image enhancement lacks an objective quality measure, and methods are problem-oriented. This paper reviews spatial-domain techniques, especially point processing and histogram processing, to explain commonly used enhancement algorithms.
Problem
Image enhancement is not well defined because image quality lacks an objective measure and methods may be inadequate across problems.
Method
The paper reviews basic spatial-domain enhancement techniques, emphasizing point processing and histogram processing with mathematical explanations and algorithms.
Results
The paper provides an overview of underlying concepts and commonly used algorithms for spatial-domain image enhancement.
Takeaways & Limitations
Image enhancement methods should be selected according to the specific problem because a method effective in one case may be inadequate in another.
Takeaways & Limitations
The discussion considers gray-level images, although the same theory can be extended to color images.
Abstract
from arXiv · showhide
Principle objective of Image enhancement is to process an image so that result is more suitable than original image for specific application. Digital image enhancement techniques provide a multitude of choices for improving the visual quality of images. Appropriate choice of such techniques is greatly influenced by the imaging modality, task at hand and viewing conditions. This paper will provide an overview of underlying concepts, along with algorithms commonly used for image enhancement. The paper focuses on spatial domain techniques for image enhancement, with particular reference to point processing methods and histogram processing.
—————————— —————————— I. Introduction
Image enhancement modifies image attributes to improve interpretability for human viewers or provide better input for automated processing, with effectiveness depending on the task. The paper reviews commonly used techniques, focusing on spatial-domain methods, especially point processing and histogram processing.
- Image enhancement modifies image attributes to make an image more suitable for a specific task and observer.The selected attributes and modifications depend on the task.
- Spatial-domain techniques manipulate image pixels directly, whereas frequency-domain techniques process Fourier-transformed images before inverse transformation.The paper emphasizes spatial-domain enhancement techniques.
- Image quality lacks an objective measure, so enhancement methods are often elementary, heuristic, and problem-oriented.A method effective in one case may be inadequate for another.
- The paper provides mathematical understanding and algorithms for basic enhancement techniques, particularly point processing and histogram processing.These techniques are discussed as commonly used approaches for image enhancement.
2. Point Processing Operation
Point processing operations are spatial-domain transformations whose neighbourhood consists only of the pixel itself. The section describes grey-level transformations including negative, thresholding, logarithmic, and power-law mappings.
- Point Processing Operation: Point processing operations apply a grey-level transformation to each pixel independently.They are the simplest spatial-domain operations because the neighbourhood is the pixel itself.
- Negative Transformation: Negative transformation inverts pixel values using N(r, c) = 255 – I(r, c).The result is useful for enhancing white or grey detail embedded in dark regions.
- Thresholding Transformation: Thresholding transformations isolate an object of interest from its background for segmentation.The transformation is presented as particularly useful when separating foreground objects from background regions.
- Log Transformation: Log transformation uses s = c * log (1 + r) to expand a narrow range of low input grey levels.Its inverse performs the opposite mapping, and log functions are useful when input grey levels span an extremely large range.
- Power-Law Transformation: Power-law transformation uses s = crγ, with different γ values producing different levels of enhancement.This technique, also called gamma correction, is commonly built into display monitors to adjust image intensity and clarity.
OUTPUT
The section presents point-processing and histogram-based methods for improving image appearance, including user-defined transforms, grey-level slicing, histogram equalization, matching, and local enhancement. It concludes that algorithm choice depends on computational cost and that combining methods may improve enhancement effectiveness.
- Point Processing: User-defined transforms can replace a well-defined mathematical function, while contrast stretching adds contrast to poor-quality images.
- Point Processing: Grey-level slicing, the spatial-domain equivalent of band-pass filtering, emphasizes selected intensity ranges while diminishing or preserving other grey levels.
- Histogram Processing: Histogram equalization seeks an output image with a uniform histogram, whereas histogram matching maps one image’s grayscale distribution to another’s.
- Histogram Processing: Local enhancement computes and applies histogram equalization or specification within a moving neighborhood around each pixel.
- Concluding Remarks: Computational cost can critically influence algorithm selection for real-time applications, and combining enhancement methods may produce more effective results.