Quantum and impulse noise filtering from breast mammogram images

Area of Science:

• Medical Imaging
• Image Processing
• Artificial Intelligence

Background:

• Noise significantly degrades mammographic image quality, impacting diagnostic accuracy.
• Accurate classification of mammograms is essential for early breast cancer detection.
• Existing noise reduction techniques may not sufficiently address severe image corruption.

Purpose of the Study:

• To develop and evaluate a robust noise detection and filtering method for mammographic images.
• To improve the accuracy of image classification by mitigating noise-induced errors.
• To enhance the overall quality of mammograms for better diagnostic interpretation.

Main Methods:

• A two-module approach: noise detection using a neural network and subsequent noise filtering.
• Neural network trained on pixel values and other features to identify noise in corrupted images.
• Noise removal employs a weighted average of three distinct filters applied to noisy pixels.

Main Results:

• The proposed technique effectively detects noise, even in highly corrupted mammographic images.
• Tested on salt & pepper and quantum noise, demonstrating superior performance over existing methods.
• Quantitative evaluation using Peak Signal to Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) confirms improved results.

Conclusions:

• The proposed noise detection and filtering method significantly enhances mammographic image quality.
• This advancement is critical for improving the accuracy of computer-aided diagnosis in mammography.
• The technique offers a promising solution for handling noisy mammograms, leading to more reliable classifications.