Histogram equalization using a selective filter

Area of Science:

• Computer Vision
• Image Processing
• Digital Signal Processing

Background:

• Popular image processing software often uses a naive histogram equalization method, resulting in non-uniform histograms.
• Existing exact histogram equalization techniques can introduce undesirable artifacts.
• There is a need for improved histogram equalization methods that balance uniformity with artifact avoidance.

Purpose of the Study:

• To bridge the gap between continuous theory and discrete implementation of global histogram equalization.
• To develop a novel histogram equalization technique that improves upon the naive approach.
• To achieve a more uniform histogram with preserved intensity distances and high entropy.

Main Methods:

• Formulated a novel histogram equalization technique based on continuous theory.
• Employed linear interpolation of the cumulative distribution for low-bit images.
• Used selective box filtering for approximate dequantization of intensities.

Main Results:

• The proposed method produces an equalized histogram with high entropy.
• Distances between similar intensity values are preserved.
• The technique offers improvements over existing naive and exact histogram equalization methods.

Conclusions:

• The novel histogram equalization technique effectively enhances image contrast and uniformity.
• The method avoids artifacts associated with some exact techniques.
• This approach has potential applications in related image processing tasks like edge detection.