Performance Evaluation of Different Techniques for texture Classification
This work addresses texture classification for image processing applications, but it is incremental as it compares existing methods without introducing new techniques.
The paper compared texture classification methods, specifically wavelet transforms (Haar, Symlets, Daubechies) and co-occurrence matrices, evaluating time complexity and accuracy. It found that the Haar wavelet was the most efficient overall, though co-occurrence matrices provided excellent accuracy except when images were rotated.
Texture is the term used to characterize the surface of a given object or phenomenon and is an important feature used in image processing and pattern recognition. Our aim is to compare various Texture analyzing methods and compare the results based on time complexity and accuracy of classification. The project describes texture classification using Wavelet Transform and Co occurrence Matrix. Comparison of features of a sample texture with database of different textures is performed. In wavelet transform we use the Haar, Symlets and Daubechies wavelets. We find that, thee Haar wavelet proves to be the most efficient method in terms of performance assessment parameters mentioned above. Comparison of Haar wavelet and Co-occurrence matrix method of classification also goes in the favor of Haar. Though the time requirement is high in the later method, it gives excellent results for classification accuracy except if the image is rotated.