Improvement and application of the orientational filter for texture image analysis

The purpose of this work is to improve the orientational filter for texture image analysis. It consists of four bandpass Quadrature Polar Separable (QPS) filters with the same bandwidth but different orientation. The frequency response of the designed QPS filter consists of the independent pair set. The first is the same as the radial weighting function used in the Knutsson's filter except inserting the shift parameter to the central frequency. The second, the orientational weighting function, is characterised by a quadrature function pair which are generated by combining an exponential attenuation function. It is easy to control the filter property in frequency domain and is more effective to estimate texture image. In order to estimate the orientations and the frequency components of local texture in frequency domain, a series of experiments have been carried out. The results show that it can be used as an efficient tool in texture processing.