China 
Africa 
Explore chapters and articles related to this topic
Image Classification and Retrieval
Published in R. Suganya, S. Rajaram, A. Sheik Abdullah, Big Data in Medical Image Processing, 2018
R. Suganya, S. Rajaram, A. Sheik Abdullah
Jero (Abd-Elmoniem et al. 2002) proposed an approach to perform ECG steganography using curvelet transform. That is, using the transform domain technology to perform steganography. He used the MIT-BIH normal sinus rhythm database of ECG signals for the experiment purpose. Here the Fast Discrete Curvelet Transform has been used for the conversion of 1D signal to 2D signal and curvelet coefficients are calculated. Patient’s data converted into binary form and threshold selection has been performed to select the coefficients (that are not representing the curve) and the n*n sequence method has been applied to the coefficients that are selected, to avoid overlapping. This method leads to the better extraction of the watermark bits. Inverse of the watermark embedding technique is used to extract the watermark bits. BER (Bit Error Rate) is observed to be zero for all the test cases which proves that embedded patient data can be extracted without any loss. PSNR (Percentage Signal to Noise Ratio) value of 75% has been observed for the small data size. And it is decreased at a rate of 10% for 1.5 increase of data size.
Noise Removal Techniques in Medical Images
Published in Siddhartha Bhattacharyya, Anirban Mukherjee, Indrajit Pan, Paramartha Dutta, Arup Kumar Bhaumik, Hybrid Intelligent Techniques for Pattern Analysis and Understanding, 2017
In this chapter, we proposed a new method for medical image denoising which is based on curvelet transform using the concept of thresholding functions combined with the Wiener filter. Initially, a noisy image is acquired with various types of noises with different noise factors. Then, threshold estimation is done followed by the application of the discrete curvelet transform to the acquired noisy image. We apply the wrapping method to implement the curvelet transform to obtain curvelet coefficients of distinct scale and orientations. The curvelet transform decomposes the acquired noisy image into different sub-bands; each sub-band is partitioned into different sub-blocks, the partitions are smoothed, the image is renormalized, and the ridgelet transform is applied. Then we apply the computed threshold value to the curvelet coefficient. After that we apply the inverse transform on the noisy image to get the denoised image. The residual noise, if any, is eliminated by the application of the Wiener filter. A quantitative estimation of the different parameters is done to evaluate the quality of the resulting denoised image. We found that the curvelet transform helps to overcome the problem of the wavelet transform, it also preserves the edge information. The curvelet transform process removes most of the noises present in the acquired image. Also, remaining residual noises are removed using the Wiener filter.
Image Processing Concepts
Published in Manas Kamal Bhuyan, Computer Vision and Image Processing, 2019
The Contourlet transform contourlet transform is proposed to efficiently represent directional and anisotropic singularities of the images. Contourlet transform provides a flexible multi-resolution, local and directional expansion for images. Basically, like curvelet and shearlets, it is proposed to overcome the limitations of wavelets for 2D singularities. The superiority of curvelet, shearlet, and contourlet with one another is highly dependent on the data and type of computer vision tasks. However, the computational complexity of curvelet is higher as compared to shearlet and contourlet [87].
Identification of wool and cashmere fibers based on multiscale geometric analysis
Published in The Journal of The Textile Institute, 2022
Liran Zang, Binjie Xin, Na Deng
In this paper, a fiber recognition technique based on multiscale geometric analysis is proposed. Curvelet transform is a method of multiscale geometric analysis, which is used to calculate the distribution features with improved direction and has the optimal characterization ability for fiber images containing a lot of curve edge information. The multi-dimensional high and low-frequency feature vectors were obtained by Curvelet transformation, and the advantages of Alexnet neural network, such as strong parallel processing ability, accurate recognition and prediction, and good robustness, were fully utilized to carry out network training for feature vectors. The network test results show that the method proposed in this paper has better recognition accuracy and higher working efficiency. It also provides a new scale segmentation and fiber recognition method, providing reasonable and effective new research ideas for other similar fields.
Generation of enhanced information image using curvelet-transform-based image fusion for improving situation awareness of observer during surveillance
Published in International Journal of Image and Data Fusion, 2019
A wrapping-based curvelet approach is proposed for fusion of infrared and low light visible images used during ground surveillance operation. Curvelet transform operates on images in a similar manner as human visual system works and decomposes images into frequency coefficients at every scale, angle and orientation. The proposed method fuses coarse scale coefficients using a PCA-based approach and detailed coefficients using absolute maximum selection rule. The fused image results obtained by the proposed fusion approach are compared with the fused results obtained by other fusion approaches. Both visual and statistical analyses of the fused images suggest that the proposed fusion approach is able to construct a better fused image with crisp background details and distinct hot target presence. The use of proposed fusion technique for fusion of infrared and low light visible images will help in enhancing observer’s SA by providing target information and background details in one single image. The fused image output thereby enhances target detection capability of observer performing surveillance even during poor ambient lighting conditions by improving perception which is the first step to achieve optimal SA.
Enhancement in the Vision of Branch Retinal Artery Occluded Images Using Boosted Anisotropic Diffusion Filter – An Ophthalmic Assessment
Published in IETE Journal of Research, 2022
S.G. Gayathri, S. Joseph Jawhar
Curvelets are basically called to be multiscale ridgelets pooled with a spatial bandpass filtering operation to segregate different scales. Recent work shows that thresholding of discrete curvelet coefficients provide near-optimal -term representations of otherwise smooth objects with discontinuities along curves. The curvelet transform turns to be a promising potential in basic application areas such as image processing, data analysis, and scientific computing. Figure 5(a) displays the three inputs collected from the clinical database and Figure 5(b) shows the curvelet transformed image for all three input image.