NON-INVASIVE NEONATAL GOLDEN HUE DETECTOR
Species Classification and Disease Identification Using Image Processing and Convolutional Neural Networks
A novel meta-heuristic jellyfish Optimize for Detection and Recognition of Text from complex images
Rice Leaf Disease Detection Using Convolutional Neural Network
Comparative Analysis of usage of Machine learning in Image Recognition
Identification of Volcano Hotspots by using Resilient Back Propagation (RBP) Algorithm Via Satellite Images
Data Hiding in Encrypted Compressed Videos for Privacy Information Protection
Improved Video Watermarking using Discrete Cosine Transform
Contrast Enhancement based Brain Tumour MRI Image Segmentation and Detection with Low Power Consumption
Denoising of Images by Wavelets and Contourlets using Bi-Shrink Filter
Although there are a lot of advances in biomedical imaging techniques, there are several factors that lead to low contrast, less visibility, and low quality of the image due to different image acquisition techniques. It is difficult to know the exact reason for this loss in quality of an image. Hence, there is a need to enhance the quality of an image of the required data available for analysis. This may be a mandatory step for every CAD tool. In this paper, an attempt is made to generate Hardware IP Core for the best enhancement technique using VERILOG. In order to recognize the best enhancement technique, a comparative analysis between two image enhancement techniques like Histogram Equalization (HE) and Power-Law transformation is performed by taking different modalities like MRI, CT of brain and lung. Although histogram equalization achieves comparatively better performance on almost all types of images, it sometimes produces excessive visual deterioration. So, the effective adjustment of contrast over an image can be done by Power-Law transformation by varying values of gamma. It is concluded that power-law transformation gives better objective parameters like Peak Signal-to-Noise Ratio (PSNR) and Mean Square Error (MSE) when compared to HE at gamma value equal to 1.01. The novelty of the proposed paper is to generate IP Core for the power-law transformation based on gamma value. This technique is implemented in Xilinx VIVADO and the same is used for simulation and calculated hardware parameters like loop latency, number of flip flops used, area consumption, memory utilization and timing constraints.
Tuberculosis (TB) is a global health problem and an infectious disease for people with low immunity and HIV/AIDS patients. Unfortunately diagnosing tuberculosis is still a major challenge. In any medical diagnosis, classifier plays an important role. In order to improve the speed of the classifier; it is required to use the selected features. To reduce the dimensionality of feature vector the authors have used RST based Multi Kernel Fuzzy C-Means (MKFCM) method for selecting features which are indispensable. Features those are not selected are treated as redundant or superfluous and are removed from the final feature vector. The performance of proposed methodology is analyzed in terms of accuracy, performance curve and regression plots. As a result, the suggested method gives the better performance.
Haze removal is a serious problem while dealing with single image. In this paper, the authors have proposed a new simple and powerful method to dehaze an image called color attenuation prior. Here, a depth map of the image has to be created at first, from a previously created linear model under the novel prior. From this find the transmission map so as to retrieve the depth information clearly. Then the last step is scene radiance recovery from which it is possible to get the dehazed image. The scene radiance recovery is done by the using the difference between saturation and the brightness of pixels. The experimental results show that the proposed method is very efficient and has a advantage, that it can dehaze sky images too.
Magnetic Resonance Imaging (MRI) technology is used to study the internal structure of brain in the form of digital images. The accurate detection of tumor region in the brain images is a challenging task. Brain tumor detection is an important task for doctors to give better treatment for the patients. Brain tumor regions can be effectively identified and located by segmentation of MRI brain images. This paper discuss and compares the efficiency of two novel optimization methods for Detection and Segmentation of MRI brain images namely “Shuffled Frog Leaping Algorithm (SFLA)- Expected Maximization (EM) frame work” and “Shuffled Frog Leaping Algorithm (SFLA) –Tabu Search (TS) frame work” for brain tumor detection in 2D MRI brain images. The results obtained had been compared with Particle Swarm Optimization Incorporating Fuzzy C Means Clusrering (PSO-FCM) method and EM methods. Finally, the results show that SFLA-TS method gives better results when compared to SFLA-EM method in identifying tumor regions in 2D MRI brain images.
Image compression is an important methodology to compress different types of images. In modern days, as one of the most fascinating machine learning techniques, the authors have applied the idea of Deep Learning in different cases of Neural Networks to prove and justify that it is the most flexible method to analyze and compress the images. Different types of neural networks are available such as Deep Neural Network (DNN), Convolutional Neural Network (CNN), Binarized Neural Networks (BNN), Artificial Neural Networks (ANN) to perform image compression. So, in this review paper the authors have discussed how deep learning concept is applied on different types of Neural Networks in order to achieve image compression of perfect qualities with proper image classifications. In order to obtain that proper image classification, ther is a need for deep learning on DNN, CNN, BNN, ANN and apply the same concept in different types of images in a justified manner with difference of analysis. This is called compression technique based on conceptual analysis of images.
