Color is the Most Demonstrative Visual Feature and Studied in the Context of CBIR

Texture is psychophysically recognized in Human Visual System (HVS), especially from the viewpoint of direction and proportion of texture pattern. Texture is also an important visual feature, which refers to the original surface characteristics of the object and its relationship to the surrounding environment. Many objects in an image can only be distinguished by their texture without additional information. Among the conventional texture features for CBIR are statistical (GLCM), Markov random field (MRF) model, simultaneous autoregressive (SAR) model, Waldo decomposition model, and edge histogram descriptor There is a texture. Characteristics (EHD) etc.

A common way to display images is to analyze the colors contained in the images. Color is the most prominent visual feature in CBIR because it is closely related to human visual perception of objects in the image. A digital color image is represented as an array of pixels containing three or four color component tuples, each pixel represented in digital form. The abstract mathematical representation of colors that a computer can use is called a color model. Similarity between images and videos is calculated using the values ​​in the color histogram. A histogram represents a specific value of a pixel in an image or video frame. Current color based search techniques divide an image into regions by using a color scale. Color based technology does not depend on image size and orientation

The CBIR system extracts image descriptors using image processing and image transformation. The CBIR system is based on different image feature descriptor matching. Some of these systems perform image comparison at the same time through multiple functions, and some of these systems use only one function. In this paper, we study basic techniques used in CBIR system based on different feature descriptors. We will explain these basic methods in detail. It also implements one of the most effective algorithms in the CBIR field. This is a scale invariant feature transformation (SIFT) algorithm (Lowe, 2004) and see how effective and accurate it is.

Existing CBIR techniques can generally be classified according to the features they are using for searching (ie color, shape, texture, or a combination thereof). Color is a widely used visual attribute that plays an important role in searching for similar images. It has been observed that even if color plays a decisive role in image retrieval, it has better results when combined with other visual attributes. This is because two images with perfectly similar color configurations may have different color configurations, as shown in the figure, because the two images may have the same color configuration and are not similar. Therefore, similar things are not semantically similar. The color composition of the two images in Figure 11 are the same, but they represent completely different meaning concepts. By analyzing two images using color-based search technology, the two images are similar