Optimal filter bank representation in texture classification using support vector machines

Texture analysis has been a field of study for over three decades in many fields including electrical engineering. Today, texture analysis plays a crucial role in many tasks ranging from remote sensing to medical imaging. Researchers in this field have dealt with many different approaches, all trying to achieve the goal of high classification accuracy. The main difficulty of texture analysis was the lack of ability of the tools to characterize adequately different scales of the textures effectively. The development in multi-resolution analysis such as Gabor and Wavelet Transform help to overcome this difficulty. This thesis describes the texture classification algorithm that uses the combination of statistical features and co-occurrence features of the Discrete Wavelet Transformed images. The classification accuracy is increased by using translation-invariant features generated from the Discrete Wavelet Frame Transform. The results are further improved by focussing on the transformed images used for feature extraction by using filters which essentially extract those areas of the image that discriminate themselves from other image classes. In effect, by reducing the spatial characteristics of images that contribute to the features, the texture classification method still has the ability to preserve the classification accuracy. Support Vector Machines has proved excellent performance in the area of pattern recognition problems. We have applied SVMs with the texture classification method described above and, when compared to traditional classifiers, SVM has produced more accurate classification results on the Brodatz texture album.

Pneumonia Detection Using Deep Learning Architectures

Pneumonia is an infection caused by bacteria and viruses. It can shift from mellow to serious cases. This disease causes severe damages to the lungs since they fill with fluids. This situation causes difficulty in breathing. It further prevents oxygen to reach the blood. Pneumonia is diagnosed with the help of a chest X-rays, which can also use in the diagnosis of diseases like emphysema, lung cancer, and tuberculosis. According to WHO (World Health Organization (WHO). 2001. Standardization of Interpretation of Chest Radiographs for the Diagnosis of Pneumonia in Children. p.4.), Chest X-rays, at present, is the best available method for detecting pneumonia. Feature extraction methods like DiscreteWavelet Transform (DWT),Wavelet Frame Transform (WFT), andWavelet Packet Transform (WPT) can be used followed by any classification algorithm. In this paper, models like Squeezenet, DenseNet, and Resnet34 have been used for image recognition. In our system, the medical images were taken from Kaggle database and were recorded using a suitable imaging system. The images retrieved were then considered for input for the system where the images go through the various phases of image processing like pre-processing, edge detection and feature extraction. Later, a variety of training models are applied to know which model offers the highest accuracy.

Content-based image retrieval using a signature graph and a self-organizing map

In order to effectively retrieve a large database of images, a method of creating an image retrieval system CBIR (contentbased image retrieval) is applied based on a binary index which aims to describe features of an image object of interest. This index is called the binary signature and builds input data for the problem of matching similar images. To extract the object of interest, we propose an image segmentation method on the basis of low-level visual features including the color and texture of the image. These features are extracted at each block of the image by the discrete wavelet frame transform and the appropriate color space. On the basis of a segmented image, we create a binary signature to describe the location, color and shape of the objects of interest. In order to match similar images, we provide a similarity measure between the images based on binary signatures. Then, we present a CBIR model which combines a signature graph and a self-organizing map to cluster and store similar images. To illustrate the proposed method, experiments on image databases are reported, including COREL,Wang and MSRDI.

A New Multi-Focus Image Fusion Algorithm Based on Contrast Ratio and Discrete Wavelet Frame Transform

This paper proposes a multi-focus image fusion algorithm based on contrast ratio and discrete wavelet frame transform. Firstly, this algorithm uses wavelet transform to perform the wavelet decomposition of the source image, and then obtains the high-frequency sub-band coefficients after the discrete wavelet frame transform to reflect the details of the image, finally, gets the fusion image obtained by wavelet reconstruction. Using evaluation indicators of information entropy, standard deviation, average gradient and spatial frequency, it objectively evaluates the fusion quality of this algorithm. The experimental results show that the quality and effect of the fusion image derived from the algorithm are significantly improved.