Madhu S. Nair

An Improved Decision-Based Algorithm for Impulse Noise Removal

The paper proposes an improved fast and efficient decision-based algorithm for the restoration of images that are highly corrupted by Salt-and-Pepper noise. The new algorithm utilizes previously processed neighboring pixel values to get better image quality than the one utilizing only the just previously processed pixel value. The proposed algorithm is faster and also produces better result than a Standard Median Filter (SMF), Adaptive Median Filters (AMF), Cascade and Recursive non-linear filters. The proposed method removes only the noisy pixel either by the median value or by the mean of the previously processed neighboring pixel values. Different images have been tested by using the proposed algorithm (PA) and found to produce better PSNR and SSIM values.

A new fuzzy-based decision algorithm for high-density impulse noise removal

This paper proposes a new efficient fuzzy-based decision algorithm (FBDA) for the restoration of images that are corrupted with high density of impulse noises. FBDA is a fuzzy-based switching median filter in which the filtering is applied only to corrupted pixels in the image while the uncorrupted pixels are left unchanged. The proposed algorithm computes the difference measure for each pixel based on the central pixel (corrupted pixel) in a selected window and then calculates the membership value for each pixel based on the highest difference. The algorithm then eliminates those pixels from the window with very high and very low membership values, which might represent the impulse noises. Median filter is then applied to the remaining pixels in the window to get the restored value for the current pixel position. The proposed algorithm produces excellent results compared to conventional method such as standard median filter (SMF) as well as some advanced techniques such as adaptive median filters (AMF), efficient decision-based algorithm (EDBA), improved efficient decision-based algorithm (IDBA) and boundary discriminative noise detection (BDND) switching median filter. The efficiency of the proposed algorithm is evaluated using different standard images. From experimental analysis, it has been found that FBDA produces better results in terms of both quantitative measures such as PSNR, SSIM, IEF and qualitative measures such as Image Quality Index (IQI).

Directional switching median filter using boundary discriminative noise detection by elimination

We propose an accurate and efficient noise detection algorithm for impulse noise removal, called the boundary discriminative noise detection by elimination (BDNDE), which retains the good characteristics of the BDND filter proposed by Ng and Ma (in IEEE Trans. Image Process. 15(6):1506–1516, 2006) while suppressing noise effectively. In order to determine whether a pixel is corrupted, the algorithm first sets the minimum and maximum boundary (threshold) values based on the localized window centered on the pixel. The thresholding helps in achieving low false-alarm and miss-detection rate (even in random noise), even up to 90% noise densities. Extensive simulation results, conducted on gray scale images under a wide range (from 10 to 90%) of noise corruption, clearly demonstrate that our enhanced switching median filter gives better results compared to existing BDND median-based filters, in terms of suppressing impulse noise while preserving image details. The proposed method is algorithmically simple and faster, compared to existing BDND, and more suitable for real-time implementation and application. The new method has shown superior performance in terms of subjective quality in the filtered image as well as objective quality in the peak signal-to-noise ratio (PSNR) measurement to that of the BDND filter.

Automatic Contrast Enhancement for Low Contrast Images: A Comparison of Recent Histogram Based Techniques

In this paper we compare two recent methods for automatic enhancement of the contrast of the image, based on the principle of transforming the skewed histogram of the original image into a uniform histogram. The histogram based gray level grouping (GLG) method and its variants (after Chen et al., 2006) and the fuzzy logic method (after Hanmandlu and Jha, 2006) are evaluated on three different images (gray scale as well as color) in order to ascertain which of the algorithms are better suited across a variety of images from different sensors and having varying characteristics. Based on the visual quality and the Tenengrad criterion we conclude that the FastHSV variant of the GLG method may be applied for automatic contrast enhancement across a wide variety of images.

Direction based adaptive weighted switching median filter for removing high density impulse noise

Restoration of images corrupted by impulse noises is a very common problem in the image processing. An efficient direction based adaptive weighted switching median filter for restoration of images corrupted by high density impulse noise is proposed in this paper. The filtering process consists of two phases - detection phase followed by a filtering phase. The detection phase in the proposed method uses HEIND algorithm put forward by Fei Duan et al. After detecting noisy pixel positions in the image, filtering algorithm is applied to those detected pixels. All uncorrupted pixels in the image are left unchanged by the filtering algorithm. The filtering algorithm uses uncorrupted pixels in the selected four directions for taking a decision on the filtering window size. Noisy pixels are replaced by a weighted median value of uncorrupted pixels in the filtering window or directional filtering window. The weight value assigned to each uncorrupted pixels depends on its closeness to the central corrupted pixel in the current filtering window.

Fuzzy logic-based automatic contrast enhancement of satellite images of ocean

In this paper, we evaluate the conventional contrast enhancement techniques [histogram equalization (HE), adaptive HE] and the recent gray-level grouping method and the fuzzy logic method in order to find out which of these is well suited for automatic contrast enhancement for satellite images of the ocean, obtained from a variety of sensors. All the techniques evaluated were based on the principle of transforming the skewed histogram of the original image into a uniform histogram. The performance of the different contrast enhancement algorithms are evaluated based on the visual quality and the Tenengrad criterion. The inter comparison of different techniques was carried out on a standard low-contrast image and also three different satellite images with different characteristics. Based on our study, we advocate that a modified fuzzy logic method elucidated in this paper is well suited for contrast enhancement of low-contrast satellite images of the ocean.

Edge preserving single image super-resolution with improved visual quality

Abstract Super-resolution is a widely used technique to increase the resolution of an image by algorithmic methods. Super-resolution from a single image is required in many real world applications. But it is a challenging task to preserve the local edge structures and visual quality in single image super-resolution. Conventional as well as advanced methods maintain the quantitative measures, but most of the times they fail to preserve edges and visual quality. We propose here a single image super-resolution algorithm which preserves the edges and at the same time maintains the visual quality, in a relatively better way. In this method, self-examples are created from a high frequency layer which is formed by performing the difference operation between the given low-resolution image and down-scaled and subsequently up-scaled version of the low-resolution image. The proposed method computes the root mean square difference of features extracted from high frequency layers of low-resolution, interpolated high-resolution and partially reconstructed high-resolution images. These difference values are fed into a Gaussian function to compute the weights which are subsequently used to perform the weighted average. The experimental analysis proves the ability of the method in improving the visual quality as well as in preserving edge information.

Predictive-based adaptive switching median filter for impulse noise removal using neural network-based noise detector

A predictive-based adaptive switching median filter for impulse noise removal using neural network-based noise detector (PASMF) is presented. The PASMF has a noise detector stage and a noise filtering stage. The noise detector implemented using feed forward neural network detects impulse noises in the corrupted image. The filter is a modified median filter, which removes detected impulse noise from the image. In contrast to the standard median filter, the PASMF computes the median value after predicting the appropriate values for neighboring corrupted pixels of the current central pixel in the filtering window. The results show that the PASMF gives better performance visually as well as in terms of different performance measures.

Automatic segmentation and classification of mitotic cell nuclei in histopathology images based on Active Contour Model

Segmentation accuracy determines the success or failure of computerized analysis procedure in biomedical applications. This paper aims to develop a unique segmentation technique to identify mitotic nuclei from microscopy images of breast histopathology slides. The process involves detection and classification of cell nuclei based on computed features. The proposed method uses Active Contour Model for segmentation of cell nuclei and two versatile classifiers such as Support Vector Machine (SVM) and Random Forest (RF) for classification stage. Segmentation stage provides an accuracy of 95% for cell nuclei. This technique uses a single color channel and a reduced feature set for the whole process. Classification performance is evaluated in terms of sensitivity, specificity, accuracy and F-score measures. Analysis results showed good detection accuracy for RF classifier compared to SVM.

Improved BTC algorithm for gray scale images using k-means quad clustering

With images replacing textual and audio in most technologies, the volume of image data used in everyday life is very large. It is thus important to make the image file sizes smaller, both for storage and file transfer. Block Truncation Coding (BTC) is a lossy moment preserving quantization method for compressing digital gray level images. Even though this method retains the visual quality of the reconstructed image it shows some artifacts like staircase effect, etc. near the edges. A set of advanced BTC variants reported in literature were analyzed and it was found that though the compression efficiency is increased, the quality of the image has to be improved. An Improved Block Truncation Coding using k-means Quad Clustering (IBTC-KQ) is proposed in this paper to overcome the above mentioned drawbacks. A new approach of BTC to preserve the first order moments of homogeneous pixels in a block is presented. Each block of the input image is segmented into quad-clusters using k-means clustering algorithm so that homogeneous pixels are grouped into the same cluster. The block is then encoded by means of the pixel values in each cluster. Experimental analysis shows an improvement in the visual quality of the reconstructed image with high Peak Signal-to-Noise Ratio (PSNR) values compared to the conventional BTC and other modified BTC methods.