Susanta Mukhopadhyay

A fast texture segmentation scheme based on active contours and discrete cosine transform

Abstract This paper presents a novel and efficient parametric active contour model based on the texture describing properties derived from coefficients of the two-dimensional Discrete Cosine Transform (2-D DCT) for segmenting texture objects against complex backgrounds. Block based 2-D DCT is applied in the local neighbourhood of all the contour control points and some randomly chosen object points. DCT coefficients in the horizontal, vertical and diagonal directions which represent the dominant textures are used to construct the histograms those exhibit different structures for different texture regions. Heterogeneity of these histograms corresponding to contour and object points is measured by Wasserstein distance metric. Finally, DC coefficient which represents the energy of the corresponding block modulated by the calculated Wasserstein distance is employed as the external energy of the proposed active contour model. Experiments on synthetic and natural texture images show the efficiency of our model in terms of accuracy and speed.

Motion detection using block based bi-directional optical flow method

Display Omitted Optical flow based moving object detection algorithm is proposed.Bi-directional optical flow field is used for motion estimation and detection.A histogram and plot based thresholding scheme is employed for motion detection.Foreground is detected using morphological operation, connected component analysis.Our technique is compared with existing methods using real video datasets. Detecting moving objects from video frame sequences has a lot of useful applications in computer vision. This proposed method of moving object detection first estimates the bi-directional optical flow fields between (i) the current frame and the previous frame and between (ii) the current frame and the next frame. The bi-directional optical flow field is then subjected to normalization and enhancement. Each normalized and enhanced optical flow field is then divided into non-overlapping blocks. The moving objects are finally detected in the form of binary blobs by examining the histogram based thresholded values of such optical flow field of each block as well as the optical flow field of the candidate flow value. Our technique has been conceptualized, implemented and tested on real video data sets with complex background environment. The experimental results and quantitative evaluation establish that our technique achieves effective and efficient results than other existing methods.

Object segmentation using graph cuts and active contours in a pyramidal framework

Graph cuts and active contours are two very popular interactive object segmentation techniques in the field of computer vision and image processing. However, both these approaches have their own well-known limitations. Graph cut methods perform efficiently giving global optimal segmentation result for smaller images. However, for larger images, huge graphs need to be constructed which not only takes an unacceptable amount of memory but also increases the time required for segmentation to a great extent. On the other hand, in case of active contours, initial contour selection plays an important role in the accuracy of the segmentation. So a proper selection of initial contour may improve the complexity as well as the accuracy of the result. In this paper, we have tried to combine these two approaches to overcome their above-mentioned drawbacks and develop a fast technique of object segmentation. Here, we have used a pyramidal framework and applied the mincut/maxflow algorithm on the lowest resolution image with the least number of seed points possible which will be very fast due to the smaller size of the image. Then, the obtained segmentation contour is super-sampled and and worked as the initial contour for the next higher resolution image. As the initial contour is very close to the actual contour, so fewer number of iterations will be required for the convergence of the contour. The process is repeated for all the high-resolution images and experimental results show that our approach is faster as well as memory efficient as compare to both graph cut or active contour segmentation alone.

A novel texture segmentation method based on co-occurrence energy-driven parametric active contour model

In this paper, a novel approach to texture segmentation based on the parametric active contour model (ACM) is proposed. At first, gray-level co-occurrence matrix and subsequently co-occurrence energy of the regions inside and outside of the dynamic contour are calculated. Difference of this energy corresponding to both the regions is used as the external energy of the proposed ACM. The contour stops and converges completely when this difference attains a maximum value. The proposed approach requires only initial contour selection and no object point selection like the other variants of parametric ACM used for texture segmentation. Experiments on a number of synthetic and real-world texture images show that in all cases, we are getting a better segmentation of the object although for few cases the execution time is bit more than that of other existing methods.

A Morphological Color Image Contrast Enhancement Technique Using Hilbert 3D Space Filling Curve

This paper presents a method based on mathematical morphology for enhancing the contrast of a color image using total ordering with Hilbert space filling curves. The method defines a total ordering of three-dimensional (3D) space (RGB space) using Hilbert 3D space filling curve and then applies morphological operators on the color image to obtain the contrast enhanced image. The output obtained through the above method has been compared with the outputs obtained through marginal morphology and vector morphology based on a distance measure. Experimental results show the efficiency of the proposed method in terms of enhanced contrast and better time complexity.

Efficient Contrast Enhancement Based on Local–Global Image Statistics and Multiscale Morphological Filtering

In this paper, image contrast enhancement is achieved by combining together the local–global image statistics and multiscale morphological filtering (MMF). The proposed method has been executed on two different sets of images, and the result has been compared with that of some existing standard methods, namely histogram equalization (HE), contrast limited adaptive histogram equalization (CLAHE), and multiscale morphology in order to have an outlook on the relative performances. The experimental results manifest that the proposed method produced results superior to the methods compared.

Texture description using multi-scale morphological GLCM

Texture is the collective repetitive pattern that characterizes the surface of real world objects. The main challenge in the texture description is its application specific definition. The present work aims at bringing the definition of textures under a generalized framework and propose some texture descriptors. In order to accomplish this, authors have extensively studied the properties of texture, drawn four observations and used some of them to devise two texture descriptors under the framework of multi-scale mathematical morphology and co-occurrence matrices. Thereafter, the descriptors are used for texture classification and tested on three benchmark datasets. Before applying the descriptors to texture classification, a dependence between number of decomposition levels (scales) and classification percentage is established using hypothesis testing. Once the dependence is established, the corresponding scale and distance parameter is chosen for each dataset. The classification results are compared with a number of existing methods. The efficacy of results prove the supremacy of the proposed methods over the existing ones.

Object Segmentation in Texture Images Using Texture Gradient Based Active Contours

Active contour models are one of the most popular and effective models for object segmentation. These models are usually dependant on the intensity gradient of the image. However, using such a model it is not possible to segment texture objects due to local convergence problem. So, we have used texture gradient instead of the intensity gradient in our proposed active contour model for texture segmentation, which is found out using non-decimated complex wavelet transform. Experimental results show that the proposed active contour model can effectively segment texture objects from their complex backgrounds in case of synthetic as well as natural texture images.