Papers by Keyword: Content Based Image Retrieval

Abstract: Contourlet-1.3 transform has better performance in directional information representation than the original contourlet transform due to less artifacts and local frequency characteristics, and has been studied by us in retrieval systems and has been shown it is superior to contourlet ones at retrieval rate. In order to improve the retrieval rate further, a novel contourlet-1.3 transform based texture image retrieval system was proposed in this paper. In the system, sub-bands energy, standard deviation and kurtosis in contourlet domain were cascaded to form feature vectors, and the similarity measure function was Canberra distance. Experimental results show that this contourlet-1.3 transform based image retrieval system has higher retrieval rates about 7% to that of the contourlet transform with absolute mean sub-bands energy and standard deviations features under same system structure.

Abstract: Content based image retrieval (CBIR) is an essential task in many applications. Color based methods have received much attention in past years, since color could serve efficiently for image retrieval, especially in the case of large database. However, there are two main drawbacks for color based image retrieval methods. Firstly, color based methods are not suitable for similar scenes under different illumination conditions, because color is sensitive to illumination. Secondly, existing approaches usually employ image descriptors with large size, which makes the approach unsuitable for real-time application. To overcome drawbacks mentioned above, an adaptive image retrieval method has been proposed, which integrates the color invariant with the spatial information about images. Different from previous methods, the quantization of the color space has not been manually determined. Instead, it has been decided according to the content of image, using an adaptive clustering technique. Therefore, the size of image descriptor is very small. In the proposed method, feature maps for images have been firstly established, which consist of color invariants. And then the Markov chain model has been employed to capture color information and spatial features. Finally, similar images are retrieved based on two-stage weighted distance. Experimental results show that the proposed method has improved simplicity and compactness of color based image retrieval methods, without the loss of efficiency and robustness.

Abstract: To support content based image retrieval, MPEG-7 is developed to define the content interfaces for images. In MPEG-7, Dominant Color Descriptor (DCD) is considered as the most important feature, and is widely used to describe the color features of an image. To support semantic queries from users, we proposed a color feature semantic mapping method in this work, which can translate the DCD values into semantic color names. The semantic mapping method is realized by constructing a mapping table between the DCD values and the semantic color names. To validate the effectiveness of our mapping method, an image retrieval experiment is conducted. From the comparison with the manually indexed description, the proposed mapping method is proved to be effective by the experiment results. Our work is very important to automatically generate the semantic description of an image and then support the users’ semantic retrieval queries.

Abstract: Contourlet-1.3 transform has fewer artifacts than original contourlet transform proposed by Do in 2002; it can extract image texture information more efficiently and has been studied for image de-noising, enhancement, and retrieval situations. Focus on improving the retrieval rate of contourlet-1.3 transform retrieval system, a new contourlet-1.3 texture retrieval algorithm was proposed in this paper. The feature vector of this system was a combination of sub-band energy and consistency and the similarity measure function used here was Canberra distance. Experimental results on 109 texture images coming from Brodatz album show that using the new features can make a higher retrieval rate than the combination of standard deviation and energy which is most commonly used today under the same retrieval time and system structure.

Abstract: Contourlet transform has better performance in directional information representation than wavelet transform and has been studied by many researchers in retrieval systems and has been shown that it is superior to wavelet ones at retrieval rate. In order to improve the retrieval rate further, a contourlet-S transform based texture image retrieval system was proposed in this paper. In the system, the contourlet transform was constructed by non-subsampled Laplacian pyramid cascaded by critical subsampled directional filter banks, sub-bands absolute mean energy and kurtosis in contourlet-S domain are cascaded to form feature vectors, and the similarity metric is Canberra distance. Experimental results on 109 brodatz texture images show that using the features cascaded by absolute mean and kurtosis can lead to a higher retrieval rate than the combination of standard deviation and absolute mean which is most commonly used today under same dimension of feature vectors. contourlet-S transform based image retrieval system is superior to those of the original contourlet transform, non-subsampled contourlet system under the same system structure with same length of feature vectors, retrieval time and memory needed, contourlet-S decomposition structure parameters can make significant effects on retrieval rates, especially scale number.

Abstract: contourlet transform can extract image texture information more efficiently than wavelet transform and has been studied for image de-noising, enhancement, and retrieval situations, its low retrieval rate are still not satisfied due to feature extraction and other reasons. Focus on improving the retrieval rate of contourlet transform retrieval system, a new feature named variance distribution was proposed and a contourlet retrieval system was constructed in this paper. The feature vectors were constructed by cascading the energy and variance distribution of each sub-band coefficients and the similarity measure used here was Canberra distance. Experimental results show that using the new features can make a higher retrieval rate than the combination of standard deviation and energy which is most commonly used today under the same retrieval time and hardware complexity.

Abstract: Contourlet transform and wavelet have been widely used in image processing systems including texture image retrieval applications, and many literatures have reported how to construct the retrieval systems. Among them, generalize Gaussian distribution (GGD) model is a promising one. We will compare the algorithm with another one, which uses energy and standard deviation features and Canberra distance. Experimental results on 40 texture images from MIT vision database show that the latter one has higher retrieval rates for wavelet and contourlet transform retrieval systems, which indicate that the GGD model is not accurate for charactering texture feature.

Abstract: Non-Subsampled contourlet transform can extract image texture information more efficiently than basic contourlet transform and has been studied for image de-noising, enhancement, and retrieval situations, its low retrieval rate are still not satisfied due to feature extraction and other reasons. Focus on improving the retrieval rate of non-subsampled contourlet transform retrieval system, a new feature named variance distribution was proposed and then a non-subsampled contourlet retrieval system was constructed in this paper. The feature vectors were constructed by cascading the energy and variance distribution of each sub-band coefficients and the similarity measure used here was Canberra distance. Experimental results show that using the new features can make a higher retrieval rate than the combination of standard deviation and energy which is most commonly used today under the same retrieval time and hardware complexity.

Abstract: In order to improve the retrieval rate of contourlet transform retrieval system, a semi-subsampled contourlet transform based texture image retrieval system was proposed. In the system, the contourlet transform was constructed by non-subsampled Laplacian pyramid cascaded by critical subsampled directional filter banks, sub-bands standard deviation, absolute mean energy and kurtosis in semi-subsampled contourlet domain are cascaded to form feature vectors, and the similarity metric is Canberra distance. Experimental results on 109 brodatz texture images show that using the three cascaded features can lead to a higher retrieval rate than the combination of standard deviation and absolute mean which is most commonly used today under same dimension of feature vectors. Semi-subsampled contourlet transform based image retrieval system is superior to those of the original contourlet transform, non-subsampled contourlet system under the same system structure with same length of feature vectors, retrieval time and memory needed, decomposition structure parameters can also make significant effects on retrieval rates, especially scale number.

Abstract: This paper proposes a novel approach of content based image retrieval based on Log Gabor Wavelet Transform (LGWT). It is observed that LGWT better represents an image compared to Gabor Wavelet Transform (GWT). Experimental results illustrate the comparative analysis of proposed retrieval system and the retrieval system based on GWT feature descriptor. It is verified that LGWT based retrieval system improves the average precision and average recall (55.46% and 32.03% respectively) from GWT based retrieval system (50.61% and 31.63% respectively). All the experiments are performed on Corel 1000 natural image database.