MBHE TECHNIQUE : CONTRAST ENHANCEMENT ON MAMMOGRAM

Abstract

Contrast enhancement is significant for medical images like mammograms. Contrast enhancements techniques are broadly of two types: Direct and Indirect contrast enhancement technique. The most popular indirect contrast enhancement techniques are Histogram equalization (HE), Contrast limited adapted histogram equalization(CLAHE), Brightness preserving bi-histogram equalization technique(BBHE) and Recursive mean separate histogram equalization(RMSHE). Some popular direct contrast enhancement techniques are contrast stretching enhancement technique and adaptive fuzzy logic contrast enhancement technique. In this work, we have proposed a new technique for the enhancement of contrast named median-based brightness conserving bi-histogram equalization (MBHE). The experiment is conducted on standard mammogram images from the mammographic Image Analysis Society (MIAS) dataset. MIAS is a standard organization of the UK that researches mammogram images. On the basis of understanding of the mammogram image, it generates a dataset of mammogram images knows as the MIAS dataset. This dataset contains 322 files. An experimental comparison of the proposed technique is done with the most popular direct and indirect contrast enhancement. A qualitative comparison is done using metrics mean square error (MSE), signal to noise ratio (SNR), and peak signal to noise ratio (PSNR). It is observed that the proposed technique outperforms the other techniques HE, RMSHE, CLAHE, adaptive fuzzy logic contrast enhancement technique, BBHE, and contrast stretching. Along with this work, a pre-processing model for mammogram images is also proposed. This model contains two steps first is filtering and the second is contrast enhancement. In this model, first we compare all the filtering techniques. After that different contrast enhancement, techniques are compared. These experiments conducted on images from the MIMA dataset. After comparison, the best filtering and best contrast enhancement technique are proposed for the mammogram images. Here, we also proposed a new contrast enhancement technique for mammogram images named as recursive median-based histogram equalization technique (RBMHE). Qualitative comparison of all these techniques is done using three quality parameters MSE, PSNR, SNR. These results show that the proposed contrast enhancement technique gives the best result among all the contrast enhancement technique and proposed model give the best result for the pre-processing of mammogram images. Pre-processing is very efficient. It is used to increase the quality of the mammogram images. Pre-processing is the first step in the process of enhancing the quality of mammogram images. v Pre-processing is helpful in noise removal, contrast enhancement, and many other mathematical operations. The first main two-step for pre-processing is noise removal and contrast enhancement. For Noise removal from the images, several filters are developed. There are mainly five filters for noise removals such as mean filter, median filter, Gaussian filter, wiener filter, and Gaussian filter. Contrast enhancement of mammogram images is done using contrast enhancement technique for example histogram equalization (HE), contrast limited adaptive histogram equalization (CLAHE), brightness preserving bi-histogram equalization technique (BBHE) and recursive mean separate histogram equalization technique (RMSHE) and contrast stretching. In this paper, we proposed a model for the pre-processing of mammogram images. For this, a comparison of all filters is performed on different noises such as salt & pepper noise, speckle noise, and Gaussian noise. After comparison, the best filter is proposed for mammogram images. Different contrast enhancement techniques are also compared and the best contrast enhancement technique is proposed in this model. Along with the proposed model, a new contrast enhancement technique named as a recursive median-based histogram equalization technique (RMBHE) is proposed. The experiment is conducted on standard mammogram images from the mammographic Image Analysis Society (MIAS) dataset. MIAS is a standard organization of the UK that researches mammogram images. This dataset contains 322 files. Experimental comparison of this proposed technique is done with the most popular direct and indirect contrast enhancement techniques such as CLAHE, BBHE, RMSHE, CONTRAST STRETCHING, ADAPTIVE FUZZY ENHANCEMENT TECHNIQUE, and HE. A qualitative comparison is done using metrics mean square error (MSE), signal to noise ratio (SNR), and peak signal to noise ratio (PSNR). It is observed that the median filter give the best noise removal for mammogram images compare to other filters such as median filter, Gaussian filter, wiener filter, and adaptive median filter and the CLAHE technique give the best contrast enhancement compared to other contrast enhancement techniques such as HE, RMSHE, BBHE, and contrast stretching. Here we observe that the proposed technique RMSHE technique outperforms all other contrast enhancement techniques such as HE, RMSHE, CLAHE, BBHE, and contrast stretching.