Thitiporn Chanwimaluang

Performance evaluation of different inpainting algorithms for remotely sensed images

Image inpainting refers to a technique in which missing areas of an image are filled is such a way that it looks plausible to the human eye by retrieving information from the surrounding pixels. Degradation in remote sensing images is usually caused by dead pixels, noise, clouds, sensor problem or communication system problem. The aim of this paper is to evaluate the performance of different inpainting algorithms for remote sensing images. Satellite Images are tested on different inpainting algorithms and efficiency of each algorithm is evaluated on the basis of processing time, Root mean square error and peak signal to noise ratio.

Naturalness color enhancement for THEOS images

Color naturalness is crucial for visual interpretation and display of satellite images. Since optical sensors in some satellite such as THEOS do not cover the whole wavelength range of red, green and blue color spectrums, the captured images may appear to have unnatural colors. It is imperative to correct the unnatural colors in these images. To achieve this goal, we must first measure the color naturalness of satellite images. As a result, we propose a novel metric for measuring color naturalness through the satellite image color naturalness index (SICNI). Here, the SICNI is desired to measure color naturalness of satellite images based on the perception of the human visual system. We compare the SICNI with evaluation from satellite image experts, and achieve the correlation score of 0.897. Based on the proposed SICNI, we develop the color enhancement algorithm for satellite images by transformation of color vectors in the CIELUV color space such that the SICNI is maximized.

The texture classification using the fusion of decisions from different texture classifiers

The improved of texture classification accuracy by using the probability weighted combination method of three texture features extraction consist of thE0020 Gray-Level Co-occurrence Matrix (GLCM), Semivariogram Function and Gaussian Markov Random Fields (GMRFs). Five different textures images are used in the experiment. The classifier that use for classify the extracted features in this research is Support Vector Machines (SVMs). The experimental result shows that the average accuracy of the combination method with probability weight up to 95.71%, which is better than the simple combination method about 2%

Automatic orchid snail detection for computed tomography images

In this paper, we address the problem of orchid snail detection from computed tomography (CT) images. Orchid snails in a CT image are barely visible since they are well-blended with the background. Moreover, noise and artifacts generated by the CT scanner are spread across the image. Therefore, the orchid snails are hardly detectable. To cope with the indistinct orchid snails, we present a new automatic orchid snail detection algorithm. First, Anisotropic diffusion technique is employed to enhance boundaries, and at the same time, try to eliminate noise and artifacts. Then, strong edges are detected by using Canny method. Subsequently, circular Hough transform is exploited to indicate the locations of orchid snails because the shape of an orchid snails boundary is quite round. Simulation results demonstrate that the proposed method can provide robust and feasible orchid snail locations.

Automatic tree-ring mark detection for teak and pine trees

In this paper, we address the problem of tree-ring mark detection for teak and pine trees. Tree ring marks, especially from teak images, are barely visible since they are well-blended with the background. Moreover, wood patterns are spread across the timber samples. In addition, noise appears all over the samples. Therefore, the tree ring marks are hardly detectable. To cope with the indistinct tree-ring marks, we present a new automatic tree-ring mark detection algorithm. First, Anisotropic diffusion technique is employed to enhance boundaries, and at the same time, try to eliminate noise and wood patterns. Then, strong edges are detected by using Canny method. Next, convolutional kernels are applied to extract horizontally-inclined lines. Subsequently, connected component labeling is used to get rid of the remaining noise. After that, mathematical manipulation is exploited to get the tree ring marks. Simulation results demonstrate that the proposed method can provide robust and feasible tree-ring mark locations.

A novel land cover mapping algorithm based on random forest and Markov random field models

This paper proposes a new land cover mapping algorithm that combines the strengths of random forest (RF) with a Markov random field (MRF) model. The idea is to transform the observed data into the decision domain of weak classifiers inside an RF. Due to how RF are trained, these decisions can be considered to be independent from each others, and therefore the joint probability density function in the decision domain can be both easily and accurately estimated. For a decision vector from RF, and under an MRF model, the optimum land cover map is iteratively searched. The performances of the proposed algorithm were evaluated using a real remote-sensing image, and we found that the resulting land cover maps are more accurate than most traditional classifiers in all sizes of training samples.

A flood mapping algorithm from cloud contaminated MODIS time-series data using a Markov random field model

This paper addresses the problem of flood detection from the cloud-contaminated MODIS time-series data. Although MODIS data can provide almost daily coverage over the large area with the medium resolution. The use of MODIS data for flood mapping in the tropical regions is a challenging task due to the cloud contamination. Since the floods usually occur in the connected regions over a certain period of time, we employed the Markov random field model to characterize this property. For our experiments, the classification accuracy of flooded and non-flooded areas can be significantly increased by incorporating the MRF model.

MAP estimation of two-sided gamma random vectors in AWGN

In this work, we present new Bayesian estimator for spherically-contoured Two-Sided Gamma random vectors in additive white Gaussian noise (AWGN). This PDF is used in view of the fact that it is more peaked and the tails are heavier to be incorporated in the probabilistic modeling of the wavelet coefficients. One of the cruxes of the Bayesian image denoising methods is to estimate statistical parameters for a shrinkage function. We employ maximum a posterior (MAP) estimation to calculate local variances with Gamma density prior for local observed variances and Gaussian distribution for noisy wavelet coefficients. The experimental results show that the proposed method yields good denoising results.