Feature extraction is important part of Content Based Image Retrieval System (CBIR). Image retrieval method with single feature has limitation. Combination of two or more features increase number of computations which is crucial factor for content based image retrieval system. In this paper, we propose a method based on two layers for image retrieval. At first layer we have used color feature and texture is used at second layer. We performed experiment on two standard, publicly available databases. We also compared our algorithm with existing methods and our method performs much better than these algorithms. Today most of the rapidly developing fields like multimedia technologies, medical, communication technologies are using digital images as an information carrier. CBIR can be used effectively manage the huge image database.
Keywords: Image retrieval; Color histogram; Gabor filter; Feature extraction; two layer
[...] In section a conclusion is given PROPOSED ALGORITHM We propose a two layer based image retrieval algorithm using color and texture feature. Feature database from image database is prepared offline before querying to the system The algorithmic steps are described as follows: Step Convert image I from RGB space to HSV space. As the result, S and V 1]. Step Quantize each pixel in HSV space to 36 bins by using quantization scheme as explained in section Step Calculate the global color histogram of image I. [...]
[...] Gabor filter is widely adopted to extract texture features from the image for image retrieval [ 13] and has been shown to be very efficient. B.S Manjunath and W.Y Ma recommended Gabor texture features for retrieval after showing that Gabor features performs better than that using pyramidstructured wavelet transform features, tree-structured wavelet transform features and multiresolution simultaneous autoregressive model. TEXTURE REPRESENTATION After applying Gabor filters on the image with different orientation at different scale, we obtain an array of magnitudes: 3 E = Gm ( y ) x y Avg. [...]
[...] 8th image is not among the ground tooth images of query image Avg Prisicion Proposed Method Global Histogram GH and Texture GH + local Color Spatial 0.08 Avg Recall Figure 7 Average Precision-Recall of Wang Database Figure 5 Retrieval results using two layer system proposed in Figure 5 shows top 10 results of two layer system proposed in Precision is 9/10. 9th image is not among the ground tooth images of query image. Figure 7 shows the average precision versus average recall for James Wang database. [...]
[...] if r = v then 6h = b g In proposed method, we are extracting color and texture feature. Global color histogram used to represent the color feature of image. Mean and standard deviation of gabor filter response is used to represent texture feature. In this section we will discuss about selection of color space, color space quantization and texture feature extraction and representation. COLOR MODEL The first step to extract color features is to select an appropriate color space. [...]
[...] In order to evaluate the performance of our proposed method, the retrieval results are compared with results of the only global color histogram method, combination of global color histogram and texture as a single layer and, global color histogram and local color spatial feature. It can be seen clearly that the proposed method outperforms all. Figure 3 to Figure 6 shows top 10 images of retrieval results of different technique for same query image. First image in top-left corner is query image SIMILARITY CALCULATION We have used city distance to calculate similarity distance at both layers. [...]
APA Style reference
For your bibliographyOnline reading
with our online readerContent validated
by our reading committee
