Abstract
A noiseless image is desirable for many applications. However, this is not possible. Generally, wavelet-based methods are used to noise reduction. However, due to insufficient performance of wavelet transforms (WT) on images, different multi-resolution analysis methods have been proposed. In this study, one of them is Contourlet Transform (CT) and the Translation-Invariant Contourlet Transform (TICT) which is an improved version of CT is compared using different noises. The fundus images are taken from the DRIVE dataset and benchmark images are used. Peak Signal-to-Noise Ratio (PSNR), Mean Squared Error (MSE), Mean Structural Similarity (MSSIM) and Feature Similarity Index (FSIM) are used as comparison criteria. The results showed that TICT is better in Gaussian noisy images.
Highlights
IN RECENT years, many algorithms have been developed for image processing [1,2,3]
Five benchmark images and 40 fundus images taken from the DRIVE dataset [32] were used
Random, Gaussian and Rician noises were added to these images respectively (sigma = 5, 10, 15 for Random noise; signal-to-ratio = 3, 5, 10 for Gaussian and Rician noise)
Summary
IN RECENT years, many algorithms have been developed for image processing [1,2,3]. Digital image processing has been available in some areas, such as physics, defense industry, medicine, industrial applications, robotics, intelligent transportation systems, etc [4]. Considering the application fields, it is understood that it is used for important and sensitive tasks. In the process of obtaining an image, noise occurs on the image. This may be caused by the quality of the camera or the environment conditions. According to Patil [5], there is no noiseless signal
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have