The content-based image retrieval (CBIR) applications have grown their popularity in the past decade with the exponential growth in the image data volumes. The social networks have aggravated the size of image data on the internet. Social network enables everyone to upload the images of one’s choice, which becomes the reason behind aggregation of millions of images on the cyber space. It’s not possible to query these large image databases with the ordinary methods. Hence there was a strong requirement of a smart and intelligent method to discover the similar images, which has been accomplished by using the machine learning methods. In this paper, the multivariate feature descriptor method has been presented to extract the required and relevant information from the large image databases. The proposed multivariate method involves the image color and texture for the purpose of image matching to the query image (also known as a reference image). The most matching entities are returned as the final results by the image extraction method. There are four methods, which involves three singular feature and one multivariate feature-based models, have been implemented. The multivariate model has been found much stable and returned the maximum accuracy under this model.