A wavelet-based spatially adaptive method for mammographic contrast enhancement.

Area of Science:

• Medical Imaging
• Image Processing
• Signal Processing

Background:

• Minimizing image noise and optimizing contrast are crucial for accurate medical image analysis.
• Existing wavelet-based contrast enhancement methods may not adequately address noise reduction.
• Mammography requires high-quality images for effective lesion detection.

Purpose of the Study:

• To develop and evaluate a generic method for simultaneous image denoising and contrast enhancement.
• To improve the visibility of image features, especially lesions in mammograms.
• To provide a quantitative and qualitative assessment of the proposed method's performance.

Main Methods:

• A novel method based on local modification of multiscale gradient magnitudes from redundant dyadic wavelet transform.
• Spatially adaptive thresholding for denoising, estimating noise from mammogram background.
• Local linear mapping for contrast enhancement, normalizing gradient magnitude maxima.
• Reconstruction from modified wavelet coefficients.

Main Results:

• Demonstrated significant performance improvement over conventional global wavelet contrast enhancement methods.
• Achieved average contrast improvement (9.04), noise amplification (4.86), and contrast-to-noise ratio improvement (3.04) indices.
• Received the highest ranking in a pilot preference study compared to other methods.

Conclusions:

• The proposed method effectively minimizes image noise while optimizing contrast, leading to superior image quality.
• Offers significant advantages for medical image analysis, particularly in mammography for lesion detection.
• The method is robust, generic, and shows high potential for clinical integration.