Texture Classification by Combining Laws Mask with Traditional Texture Analysis Techniques

Abstract

Texture is one of the primary visual attributes used to describe patterns found in nature. Texture analysis have been received a considerable amount of attention over the last few decades as it creates the basis of most pattern recognition methods. Many approaches have been proposed in signal processing methods. Laws texture energy measure is one of the important approaches to describe the texture, and known as Laws mask method. However, the original Laws mask method is not capable of capturing texture information at multiple scales. The original Laws mask has also not been utilized to verify the robustness to illumination variations. With this entire gap in mind, the central hypothesis of this thesis is to integrate the original Laws mask approach with various existing texture analysis techniques for a richer texture description in terms of classification rates.To do so, we used Steerable pyramid, Discrete Wavelet Transform, Dual Tree Complex Wavelet Transform, Bilateral filter, and Homomorphic filter, techniques that led to methodologies with efficiency comparable to the original Laws mask technique. newlineSteerable pyramid is a linear decomposition through which an image is subdivided into a collection of subbands of various orientations. Our first method combines steerable pyramid with Laws mask. Steerable pyramid with three different approaches are integrated with Laws mask. The wavelet transform extracts directional details that capture horizontal, vertical and diagonal details of an image. In a second approach, discrete wavelet transform is combined with Laws mask. Different basis functions with three decomposition levels are used in discrete wavelet transform.