CONTOURLET TRANSFORM FOR TEXTURAL IMAGE CLASSIFICATION

INTRODUCTION

Texture analysis is still considered an interesting but challenging problem in image processing field. In computer vision, several approaches have been proposed in the past for texture analyse. Recently researchers are motivated by human version system to develop multiresolution space/scale texture models such as Gabor filter and wavelet tansforu. Gabor filters require proper tuning of filter parameters at different scales; their transformations are usually not reversible, and finally, there is a significant correlation between their texture features. Wavelet transform on the other hand, has the ability to perform local analysis fur revealing various aspects of data like trends, breakdown points, discontinuities in higher derivatives, and self-similarities. A major drawback of two-dimensional wavelets is their limited capability in capturing directional information which has a significant role in analysts of the images, including feature extraction and classification. 

To overcome this deficiency,a new family of wavelet methods that can capture the intrinsic geometrical structures such as curvelet transform and eontourlet transform. Curvelets are very successful in detecting image activities along curves, while analyzing images at multiple scales, locations, and orientations. Contourlet transform proposed by Do and Vetterli , uses a structure similar to that of curvelets, except at discrete domain. The contourlet expansion is composed of basis images oriented at various directions in multiple scales, with flexible aspect ratios which effectively capture smooth contours of images.

In many remote sensing applications such as aerial or satellite photography, and underwater acoustic imaging systems. textural images that may be acquired from the same scene but with different slope, direction. distance. noise level and illumination, should be classified consistently. It was shown that wavelet transform is suitable for this task. However, computer vision literature has paid less attention to the contourlet domain texture segmentation and classification.

CONTOURLET TRANSFROM

The contourlet transform is a new two-dimensional extension of the wavelet transform proposed by Do and Vetterli using multiscale and directional filter banks. The contourlet expansion is composed of basis images oriented at various directions in multiple scales, with flexible aspect ratio that could effectively capture smooth contours of seabed images. The contourlet transform employs an efficient tree structured Implementation which is an iterated combination of the Laplacian Pyramid (LP) for capturing the point discontinuities and the Directional Filter Bank, to gather the nearby basis functions and link point discontinuities into linear structures. Since the DFB was designed to capture the high frequency directionality of the input image and it. is poor on handling low frequency content., hence the DFB is combined with the LP, where low frequency of the input image is removed before applying DFB.