In this paper, an images similarity measure based on directional gradient angular histogram (DGAH) is proposed. Analysis and experimental results reveal that the proposed technique not only acquires a superior detection performance, but also the invariance to illumination changes and the robustness to translation, scale and rotation of the image object. In addition, the proposed technique is easily integrated within a wavelet-based image coder since the angular histogram features can be constructed directly on the wavelet coefficients decomposed using anti-symmetrical bi-orthogonal wavelets (ASBW).