A comparative study of enhancement techniques applied to noisy mammogram images: An Experimental Evaluation Based on the MIAS Dataset

Abstract

Background and objective: Mammography is a common imaging method, utilizing low-dose X-rays to detect calcifications and other tissue abnormalities. The aim of this study was to determine the most effective techniques for enhancing mammogram images to ensure accurate and reliable breast cancer detection. Methods: Mammogram images from the MIAS database were enhanced using various techniques to improve image quality for more accurate diagnostic identification. The methodology proceeded in two steps. First, artificial noise — Gaussian, quantum, salt-and-pepper, Poisson, and speckle — was purposefully added to create a controlled testing environment and replicate noisy settings; although mammogram images naturally contain some noise, this artificial addition allowed systematic evaluation of enhancement and denoising techniques. Second, contrast enhancement (histogram equalization and contrast-limited adaptive histogram equalization, CLAHE) and denoising (Gaussian, median, and Wiener filters) were applied to restore image clarity while preserving diagnostically significant information. Peak signal-to-noise ratio (PSNR) was used to evaluate effectiveness, and one-way ANOVA was used to determine the significance of each technique's effect on PSNR values. Results: The Gaussian filter achieved the best performance, with PSNR values of 66.79 dB and 46.6 dB for salt-and-pepper and Poisson noise, respectively, while CLAHE performed best for Gaussian noise (PSNR=50 dB). PSNR values were significantly affected by the enhancement technique used, according to ANOVA (F=1.08×10⁴, p <0.001). Conclusion: The Gaussian filter successfully lowers noise and greatly improves image contrast, making these techniques useful for the precise and early identification of breast cancer in mammograms.