Design and Implementation of Fractional Order Integral Filter for Denoising of Echocardiographic Images

Abstract

Denoising techniques demand the areas of analysis which are capable of suppressing noise and also preserving the edges and structure with respect to image quality metrics, visual quality assessment and clinical validation. Denoising is an essential step of preprocessing for improving the diagnostic value of the images. Fractional order calculus is a demanding area of research in various fields of image processing. Fractional Calculus based Denoising filter is proposed in the paper. Denoising is accomplished by the application of filter mask on each pixel of the noisy image. Fractional order integral filters are higher in accuracy in comparison to the integer-order integral filters. The adaptive fractional integral algorithm for image denoising is the thrust area of research in image processing and can prove to be more fruitful. The fractional order denoising algorithm requires structures of nxn filter masks of digital images. The algorithm is tested to both subjective and objective standards of perceiving ability of the human visual system. Simulations using MATLAB are carried out on real echocardiographic as well as standard images. The filter is tested on synthetic Images like Lena Image, Checkerboard image and cameraman image, only the synthetic images are corrupted by adding speckle noise of different noise variances. Further, the filter was tested on real echocardiographic images in different views; since there is no reference noise-free real echo image available as the gold standard for the performance verification of the proposed denoising filter which is only based on the results of standard images. The thrust area includes a faster implementation of the proposed method compared to other methods of denoising. The results obtained using the proposed filter performs better than other recently proposed methods, both in terms of quantitative and qualitative measures. Quantitative metrics being used are Peak signal to noise ratio (PSNR), Signal to noise ratio (SNR) and Mean square error (MSE). Further, the order of the filter is adaptive which is more significant for image processing like segmentation and feature extraction. The proposed denoising technique not only improves the quality of low contrast images but also effectively reduces speckle noise in echo and synthetic images.