International Journal of Engineering and Advanced Technology (IJEAT) ISSN: 2249 – 8958, Volume-9 Issue-3, February, 2020 205 Published By: Blue Eyes Intelligence Engineering & Sciences Publication Retrieval Number: B4208129219/2020©BEIESP DOI: 10.35940/ijeat.B4208.029320 Abstract: Image data has turned out to be a significant means of expression with the advancements of digital image processing technologies. Image capturing devices has now transformed to commodities due to smart integration with cell phones and other useful devices. Huge amount of images are getting accumulated daily in gigantic databases which requires categorization for prompt retrieval in real time. Content based image classification (CBIC) thus gained it's popularity in classifying images to their corresponding categories. Feature extraction techniques are the foundation of CBIC to represent the image data in the form of feature vectors. This work has implemented three different feature extraction techniques from spatial domain, transform domain and deep learning domain. The three different feature vectors feature vector are contrasted to investigate the robustness of descriptor definition for content based image classification Keywords: binarization, image transform, pretrained CNN, feature extraction, image classification I. INTRODUCTION Classification of images based on content of the image data has revealed remarkable growth in recent times due to improved techniques of feature extraction and robust machine learning algorithms. Availability of open source architecture of neural networks for deep learning has further leveraged the research in the aforementioned domain [1]. Prior to automated feature extraction using deep neural networks, traditional techniques of handcrafted feature extraction has contributed immensely towards progressive development of content based image classification [2]. However, feature engineering has a major role in formulating robust content based features using hand crafted techniques [3]. In contrast, feature extraction using deep learning techniques hardly require any manual intervention for revealing credible feature patterns from content based image data. This work has carried out feature extraction using three different techniques. Two of the techniques are handcrafted ones and the last one is based on pretrained convolutional neural network [4]. Feature extraction using handcrafted techniques embrace two popular methods, namely, Revised Manuscript Received on February 05, 2020 * Correspondence Author Rik Das*, PGPM – Information Technology, Xavier Institute of Social Service, Ranchi, India. Email: rikdas78@gmail.com Mohammad Arshad, MCA Department, Vinoba Bhave University, Hazaribag, India. Email: arshadnel@gmail.com Pankaj Kumar Manjhi, University Department of Mathematics, Vinoba Bhave University, Hazaribag, India. Email: 19pankaj81@gmail.com binarization technique and image transform techniques. Both the techniques are highly effective to extract meaningful descriptors from image content. Bernsen's technique of local threshold selection for image binarization is one of the familiar techniques for extraction of features by separating the foreground of the image from it's background by selection of local threshold within a given window . Another useful technique for image descriptor definition is to use fractions of transform coefficients as feature vectors for classification purpose. In this work, slant transform is used for the purpose of feature vector generation using transform technique [5]. The process of selection of effective fractional coefficient is instrumental in reducing the dimension of image features extracted using transform technique. Finally, a pretrained convolutional neural network (VGG 16) is used for extraction of content based features from image data [6]. A well known open access dataset is used as test bed for the experimentation purpose. Extracted features are evaluated for classification accuracy by means of two diverse classifiers, specifically, support vector machine (SVM) and random forest. The results have revealed superior categorization of images with descriptors extracted using VGG 16 in contrast to traditional handcrafted method. II. LITERATURE REVIEW Mainstay of CBIC is principally reliant on significance of descriptor representation. Several techniques are designed to extract features manually in which feature engineering has a significant stake. Binarized statistical features are extracted using three different filters from iris images in [7]. Magnetic resonance imaging binarization is carried out in [8] that have computed the threshold based on mean, variance, standard deviation and entropy. Extremely randomized trees are used for binarization of degraded document images in [9]. SIFT features are transformed to binary representations in [10] for fast image retrieval. Transform techniques are crucial for energy compaction in images. This attribute of the transform techniques can be well utilized to extract meaningful content based image features. Slant transform is used for feature extraction for iris image recognition [11]. A novel algorithm for feature extraction is designed in [12] using slant transform. Extended classification results are achieved using transform techniques for feature extraction in [13]. Fusion is carried out at score level and feature level using fractional coefficients of image transforms [14]. Local feature extraction of video data using Assorted Techniques for Defining Image Descriptors to Augment Content Based Classification Accuracy Rik Das, Mohammad Arshad, Pankaj Kumar Manjhi