IMAGE QUALITY MEASURE USING CURVATURE SIMILARITY Susu Yao, W. Lin, Z.K Lu, E.P. Ong, M.H. Locke and S.Q. Wu Institute for Infocomm Research, Singapore 119613 E-mail: ssyao@i2r.a-star.edu.sg ABSTRACT This paper proposes a new full-reference objective metric for image quality assessment. The reference and distorted images are decomposed into a number of wavelet subbands, in which mean curvatures and perceived error of the wavelet coefficients of two images are computed and integrated to give overall quality index. Taking structural similarity and error visibility into account, the new method can achieve high consistency with subjective evaluation compared with other metrics. Experimental results have shown the effectiveness of the proposed metric. Index Terms— Image quality assessment, surface curvature, correlation, wavelet transform, error visibility. 1. INTRODUCTION The quality of a digital image is affected by many factors. In practical image processing such as acquisition, compression, transmission and reconstruction, the visual quality of image is degraded in different degrees due to added noise or loss of image information. Usually, we compare original image and processed image to evaluate visual quality, namely full-reference quality assessment. Subjective evaluation is most accurate but it is time- consuming and inconvenient. Objective evaluation methods which can automatically predicate perceived visual quality are desirable in most practical image processing systems such as image and video coding, dynamic monitoring of image quality. A number of objective image quality assessment methods have been proposed in past few years [1,2,3], in which many efforts have gone into the investigation of error sensitivity in spatial or frequency domain. However, psychological experiments have revealed that the human visual system (HVS) has more sensitivity to the structural information variation than to the error visibility. Obvious evidence is that human eyes are less sensitive to the small mean value shift of an image than to the distortions at image discontinuities. Based on this fact, Wang [4] proposed an image quality assessment method in which one of three factors measures the structural similarity using vector correlation. However, the spatial vector correlation cannot distinguish effectively the structural variation. This paper attempts to further exploit structural similarity in wavelet bands using differential geometric information because human eyes are able to deal with many geometric deformations quickly and accurately. In 3-D space, a surface model can represent image regions including flat surfaces, piecewise-smooth curved surfaces, edges and texture. We use surface curvature similarity between reference and distorted images as a factor to measure quality degradation. On the other hand, the frequency sensitivity of human perception for the wavelet coefficients is also taken into account to compute perceived errors that are integrated with the measure of curvature similarity to give overall image quality index. The effectiveness of the proposed method is verified through evaluating an image test database. The experimental results show that the new method can give a high correlation with the subjective scores in terms of prediction accuracy, monotonicity and consistence. This paper is organized as follows. In Section 2, wavelet decomposition and error sensitivity of wavelet coefficients are described. Section 3 gives brief description of surface curvature. Section 4 presents the proposed image quality assessment method. Experimental results and comparison with other metrics are given in Section 5. Finally, conclusions are drawn in Section 6. 2. WAVELETS AND ERROR SENSITIVITY MODEL 2.1. Wavelet Decomposition The wavelet transform is one of the most powerful techniques for image processing because of its similarities to the multiple channel models of the HVS. In recent years, image quality measures have used wavelet transform to perform spatial frequency decomposition [6][7]. Figure 1 shows a two-dimensional frequency space in which four-level hierarchical wavelet subbands are obtained using DWT 9/7 biorthogonal filters. In this space, there are total 13 subbands with one low-frequency subband and 12 AC subbands. Each level in the decomposition contains an LH band, an HL band, and an HH band. Figure 1. Titling of the two-dimensional frequency space by four- level hierarchical wavelet decomposition. 2.2. Wavelet Frequency Error Sensitivity Model It has been found that human eyes have different sensitivities to the different frequency bands. In [5], the base detection thresholds LH1 HH1 HL1 LH2 HH2 HL2 LH3 HH3 HL3 LL LH4 HH4 HL4 2 π 2 π 4 π 4 π ω ω III - 437 1-4244-1437-7/07/$20.00 ©2007 IEEE ICIP 2007