Article Preview
TopIntroduction
The rapid proliferation of digital images in cyberspace has inspired research into developing efficient image content management system. In the recent decade, the rapid advancement of digital capturing devices and social media has led to a huge development of image database systems (Desai et al., 2021; Zhong et al., 2021; Desai et al., 2021). Acquiring important data from these enormous databases has stimulated the research community's desire to find effective alternatives that do not rely on textual descriptions for each image. This investigation led to the discovery of a technique for content-based picture retrieval (CBIR).
In many areas, the CBIR is utilized like crime preventing, video processing, digital albums, biodiversity, medical imaging, and other areas which need image recognition. In CBIR, the images are automatically indexed based on its low level features extracted from the image like shape and color which are only used for image retrieval (Keisham et al., 2022; Kumar et al., 2022; Chen et al., 2021). The two most critical aspects of the CBIR program's implementation are feature representation and similarity measurement.
Color-based features, texture characteristics, and shape features are among the low-level features examined for clustering and retrieving images (Putzu et al., 2020). The value of pixels in RGB, HSV, and other colour spaces is extracted and represented as the image's feature vector in color-based features (Sundararajan et al., 2019; Hu et al., 2021; Zhang et al., 2022; Sezavar et al., 2019). The content of an image cannot be fully described by colour alone. As a result, employing colour attributes individually could not reach adequate accuracy. Therefore, to compare images, texture features are necessary.
In an image, texture is one of the crucial surface property which represents the more important surface related features like clouds, tiles, bricks etc. it is described by the similarity of visual patterns. The shape identifiers do not indicate that the complete shape of the image is described, but rather that the shape of a specific section of the image is described. Shapes are also utilised for segmentation and contour detection (Sharif et al., 2019; Öztürk et al., 2020; Ashraf et al., 2020). Scaling, rotation, and Invariance for translation are the approaches utilised for the shape descriptor. Using the discrete cosine transform, wavelet transform, discrete Fourier transform or a combination of them, some relevant texture features can be retrieved from images (Bibi et al., 2020).