T. Sree Sharmila

Efficient analysis of hybrid directional lifting technique for satellite image denoising

Satellite images are often corrupted by noise in the acquisition and transmission process. While removing noise from the image by attenuating the high frequency image components, it removes some important details as well. In order to improve the visual appearance and retain the useful information of the images, an effective denoising technique is required to reduce the noise level. For denoising, many researches exploit the directional correlation in either spatial or frequency domain. However, the orientation estimation for directional correlation becomes inefficient and error prone in noised circumstances. This paper proposes a new hybrid directional lifting (HDL) technique for image denoising that involves pixel classification and orientation estimation, along with adding small amount of noise, in order to improve the performance efficiency of the technique. Experimental results show that the HDL technique improves both peak signal to noise ratio and visual quality of images with rich textures.

Efficient quality analysis of MRI image using preprocessing techniques

Image pre-processing techniques are used to improve the quality of an image before processing into an application. This uses a small neighborhood of a pixel in an input image to get a new brightness value in the output image. These preprocessing techniques are also called as filtration and resolution enhancement. The medical image quality parameters are mainly noise and resolution. The main objective of this paper is to improve the image quality by denoising and resolution enhancement. Most of the imaging techniques are degraded by noise. In order to preserve the edges and contour information of the medical images, the efficient denoising and an improved enhancement technique is required. This paper concentrates the average, median and wiener filtering for image denoising and an interpolation based Discrete Wavelet Transform (DWT) technique for resolution enhancement. The performance of these techniques is evaluated using Peak Signal to Noise Ratio (PSNR). From the results, it reveals that the efficient denoising and resolution enhancement technique is essential for image pre-processing.

Impact of applying pre-processing techniques for improving classification accuracy

Image denoising is a procedure aimed at removing noise from images while retaining as many important signal features as possible. Many images suffer from poor contrast due to inadequate illumination or finite sensitivity of the imaging device, electronic sensor noise or atmospheric disturbances. This paper proposes a hybrid directional lifting technique for image denoising to retain the original information present in the images. The primary objective of this paper is to show the impact of applying preprocessing techniques for improving classification accuracy. In order to classify the image accurately, effective preservation of edges and contour details of an image is essential. The discrete wavelet transform-based interpolation technique is developed for resolution enhancement. The image is then classified using support vector machine classifier, which is well suitable for image classification. The efficiency of the classifier is analyzed based on receiver operating characteristic (ROC) curves. The quantitative performance measures peak signal to noise ratio and ROC analysis show the significance of the proposed techniques.

Comparative Analysis of Satellite Image Pre-Processing Techniques

Satellite images are corrupted by noise in its acquisition and transmission. The removal of noise from the image by attenuating the high frequency image components, removes some important details as well. In order to retain the useful information and improve the visual appearance, an effective denoising and resolution enhancement techniques are required. In this research, Hybrid Directional Lifting (HDL) technique is proposed to retain the important details of the image and improve the visual appearance. The Discrete Wavelet Transform (DWT) based interpolation technique is developed for enhancing the resolution of the denoised image. The performance of the proposed techniques are tested by Land Remote-Sensing Satellite (LANDSAT) images, using the quantitative performance measure, Peak Signal to Noise Ratio (PSNR) and computation time to show the significance of the proposed techniques. The PSNR of the HDL technique increases 1.02 dB compared to the standard denoising technique and the DWT based interpolation technique increases 3.94 dB. From the experimental results it reveals that newly developed image denoising and resolution enhancement techniques improve the image visual quality with rich textures.

NOVEL COLOR FILTER ARRAY DEMOSAICING IN FREQUENCY DOMAIN WITH SPATIAL REFINEMENT

The main idea behind wavelet based demosaicing with spatial refinement is to reconstruct the full reso lution color image from the mosaiced image. In this study, a new effective wavelet based demosaicing algorith m for interpolating the missing color components in Bayer ’s Color Filter Array (CFA) pattern is proposed. Th is interpolation technique uses the interchannel corre lation among the high frequency subbands to determine the missing pixels in each color channel, followed by a refining step in spatial domain which uses non-ite rative technique that enforces color difference rule with fewer computations. As a result, the proposed demosaicing method yields better performance than bilinear, edg e based and subband based demosaicing methods.

Acoustic image enhancement using Gaussian and laplacian pyramid – a multiresolution based technique

Acoustic images captured by side scan sonar are normally affected by speckle noise for which the enhancement is required in different domain. The underwater acoustic images obtained using sound as a source, basically contain seafloor, sediments, living and non-living resources. The Multiresolution based image enhancement techniques nowadays play a vital role in improving the quality of the low resolution image with repeated patterns. Image pyramid is the representation of an image at various scales. In this work, a three level Gaussian and Laplacian pyramids are constructed to represent the image in different resolution. The multiscale representation requires different filters at different scales. The contrast of each image in Gaussian and Laplacian pyramids are improved by applying both histogram equalization and unsharp masking method. The sharpened images are used to reconstruct the enhanced image. The performance measure, peak signal to noise ratio proves that the unsharp masking method applied to difference images of Laplacian pyramid outperforms the other image enhancement methods.

Edge based eye-blink detection for computer vision syndrome

Living in an information age the whole earth is a small globe in our hands with the advancements of computers, smartphones etc. The usage of computers in our day-to-day activities has increased enormously leading to both positive and negative effects in our lives. The negative effects are related to health problems such as Computer Vision Syndrome (CVS) etc. Prolonged use of computers would lead to a significant reduction of spontaneous eye blink rate due to the high visual demand of the screen and concentration on the work. The proposed system develops a prototype using blink as a solution to prevent CVS. The first part of the work captures video frames using web-camera mounted on the computer or laptop. These frames are processed dynamically by cropping only the eyes. The algorithms performed on the eye-frames are direct pixel count, gradient. Canny edge and Laplacian of Gaussian (LoG). These determine the eye-status based on the threshold value and the proposed idea, the difference between upper and lower eye frames. Various experiments are done and their algorithms are compared and concluded that the proposed algorithm yields 99.95% accuracy.

EM algorithm based intervertebral disc segmentation on MR images

Image segmentation is well known in partitioning a digital image into several segments. Recent days lower back pain in human being increases and so the lumber spine pathology detection becomes a predominant research area in Computer Aided Diagnosis (CAD) system. In the process of lumbar spine pathology detection, the segmentation of the Intervertebral Disc (IVD) is the major step as it identifies the IVDs or the boundaries of the IVDs either normal or abnormal in images. When the axial or the sagittal View of lumbar spine MR image is given as input, this proposed work segments the IVD in both the axial and sagittal views. The segmentation of IVD is a four stage process. First, Expectation-Maximization (EM) segmentation is performed on the MR Image. EM segmentation yields an advantage over K-means with the case of the size of clustering. The second stage is to carry out the morphological operators and third, apply edge detection method and obtain the edges. The final stage is to remove unwanted objects from the obtained output image. If this proposed segmentation is utilized as part of the CAD, the experts will be benefited for localizing the IVD and to diagnose the IVD disease.

An efficient framework for image retrieval using color, texture and edge features

Abstract This paper proposes a new hybrid framework for Content-Based Image Retrieval (CBIR) system to address the accuracy issues associated with the traditional image retrieval systems. The proposed framework initially selects pertinent images from a large database using color moment information. Subsequently, Local Binary Pattern (LBP) and Canny edge detection methods are used to extract the texture and edge features respectively, from the query and resultant images of the initial stage of this framework. Then, the Manhattan distance information about these two features corresponding to the query and selected images are calculated and combined, and then sorted using bubble sort algorithm. Wangs, Corel-5K and Corel-10K are the three databases used for evaluating the performance of the proposed hybrid framework using precision and recall measures. The average precision measured on these three databases gives approximately 11.8%–22.315%, 8.025%–18.935% and 10.755%–32.221% higher accuracy than the state-of-the-art techniques.

Text frame classification and recognition using segmentation technique

Text detection techniques assume that the frames given to them are all text frames. When a non-text frame is fed to the text detector there is an increase in the number of false positives which is a major issue in text detection and recognition. Hence, classification of text frames in a video sequence is important. The duplicate frames are eliminated and label each frame as text and non-text before it can be given for detection and recognition of text in a frame. Text in a frame is detected using bounding box technique. Once the text regions are marked, Fuzzy C Means (FCM) algorithm is used to classify each frame as text and non-text. The performance of the proposed work is measured in terms of precision, recall and F-measure and is compared with the existing approaches. The classified text frames are taken as input and are sent to the OCR Engine for character recognition. Several applications such as character recognition, video content analysis, and content based image retrieval, video summarization, and script identification etc., uses text detection.