Integrated features of haar-like wavelet filters

In this paper a novel feature descriptor by integration of cooccurrence of Haar like wavelet filter (CHLWF) with color histogram (CH) is proposed for content based image retrieval (CBIR). The proposed feature is capable of extracting image properties from different visual perspectives in order to give image representation almost similar to human interpretation. It helps in the improvement of retrieval results in terms of various performance measures. An effort is made to reduce the sensitivity to noise and illumination changes by working on average intensities of the regions. CHLWF is extracted from images and employs only maximal edge responses in feature computation hence, only prominent directional information which can substantially represent an image is used. It provides textural information efficiently with reduced computational complexity. CHLWF also eliminates the step involved in the decision making of thresholds by automatic quantization of coefficients. All these properties assist to show the efficient image representation through CHLWF. Further with the integration of CH feature color information is also incorporated in the final feature and results are improved with respect to related state of art techniques. Since, the main aim of introducing this feature is to improve effectiveness of retrieval system hence; it is affirmed by evaluating the retrieval performance on Corel 1000 benchmark image database.