International Journal of Engineering Trends and Technology (IJETT) – Volume 19 Number 2 – Jan 2015 ISSN: 2231-5381 http://www.ijettjournal.org Page 85 Gaussian DBC Co-Occurrence Matrix for Content-Based Image Indexing and Retrieval K. Prasanthi Jasmine #1 , P. Rajesh Kumar *2 , K. N. Prakash 3 # 1, * 2 Department of Electronics and Communication Engineering, Andhra University, Visakhapatnam, Andhra Pradesh, India 3 Department of Electronics and Communication Engineering, Lakireddy Balireddy College of Engineering, Andhra Pradesh, India Abstract— This paper presents a novel image indexing and retrieval algorithm using Gaussian multi-resolution directional binary code (DBC) co-occurrence matrix. DBC histogram captures only the patterns distribution in a texture while the spatial correlation between the pair of patterns is gathered by DBC Co-occurrence. Multi- resolution texture decomposition and co-occurrence calculation has been efficiently used in the proposed method where multi-resolution texture images are computed using Gaussian filter for collection of DBCs from these particular textures. Eventually, feature vectors are constructed by making into play the co-occurrence matrix that exists between binary patterns. The retrieval results of the proposed method have been tested by conducting two experiments on Brodatz and MIT VisTex texture databases. The results after being investigated show a significant improvement in terms of their evaluation measures as compared to LBP, DBC and other transform domain features. Keywords— Multi-resolution features; Gaussian Filter; Directional Binary Code; Texture; Pattern Recognition; Feature Extraction; Local Binary Patterns; Image Retrieval. I. INTRODUCTION With the growth in technology and advancement of the living world, there has been an expansion of digital database in order to meet ones’ technical requirement. Handling of these databases by human annotation is a cumbersome task thereby, arousing a dire need for some familiar search technique i. e. content based image retrieval (CBIR). The feature extraction forms a prominent stair in CBIR and its effectiveness relies typically on the method of features extraction from raw images. The visual contents of an image such as color, texture, shape, faces, and spatial layout etc. are the key pillars for representing and indexing. The visual features can further regimented into general features that includes color, texture, shape and domain specific features such as human faces and finger prints. Preeminent representation of an image for all perceptual subjectivity still does not exist as the user may take photographs under different conditions (view angle, illumination changes etc.) there by making the learning of high level semantic concepts an utmost task for CBIR systems. Comprehensive and extensive literature survey on CBIR is presented in [1]– [4]. Texture analysis has been an eye catcher due to its potential values for computer vision and pattern recognition applications. The selective visual attention model (SVAM) is incorporated in [5] for the CBIR task to estimate user’s retrieval concept. It distinguishes itself from the existing learning based retrieval algorithms as they need relevance feedback strategy to get user’s high- level semantic information. Also, an improved saliency map computing algorithm based on the saliency map, an efficient salient edges and regions detection is proposed. The dominant set clustering (DSC) similarity for image retrieval can be seen in [6] where the low-level visual features and high-level concepts are amalgated using DSC similarity measure for the relevance feedback based image retrieval system. Retrieval performance is tested on an image retrieval system using the memorized support vector machine (SVM) relevance feedback. An efficient histogram oriented gradients (HOG) based human detection [7] reduces the features in blocks for constructing the HOG features for intersection detection windows and utilizes sub-cell based interpolation that efficiently computes them for each block. Texture is another salient and indispensable feature for CBIR. Vo et al. [8] proposed the Vonn distribution of relative phase for statistical image modeling in complex wavelet domain. A relative phase probability density function, known as Vonn distribution, in complex wavelet domain and the maximum-likelihood method is used for estimating two Vonn distribution parameters. The rotation-invariant texture retrieval using wavelet- based hidden Markov trees can be seen in [9]. The feature extraction of the texture is performed by employing the signature of the texture, generated from the wavelet coefficients of each subband across each scale and the similarity that exists between textures using Kullback– Leibler (KL) distance measure. Smith et al. used the mean and variance of the wavelet coefficients as texture features for CBIR [10]. Moghaddam et al. proposed the Gabor wavelet correlogram (GWC) for CBIR [11, 12]. Ahmadian et al. used the wavelet transform for texture classification [13]. Moghaddam et al. introduced new algorithm called wavelet correlogram (WC) [14]. Saadatmand et al. [15, 16] improved the performance of WC algorithm by optimizing the quantization thresholds using genetic algorithm (GA). Birgale et al. [17] and Subrahmanyam et al. [18] combined the color (color histogram) and texture (wavelet transform) features for CBIR. Subrahmanyam et al. proposed correlogram algorithm for image retrieval using wavelets and rotated wavelets (WC+RWC) [19]. Ojala et al. proposed the local binary patterns (LBP) for texture description [20] and these LBPs are converted to rotational invariant for texture classification [21]. pietikainen et al. proposed the rotational invariant texture classification using feature distributions [22]. Ahonen et al. [23] and Zhao et al [24] used the LBP operator facial