Content-based image retrieval is a technique for locating images in vast, unlabeled image collections (CBIR). However, users are not happy with the traditional methods of information retrieval. Additionally, the number of consumer-accessible pictures and online production and distribution channels are expanding. Consequently, permanent and widespread digital image processing occurs across numerous industries. As a result, acquiring quick access to these big image databases and extracting identical images from sizable groups of photographs from a specific image (Query) create significant problems that call for efficient solutions. Calculations related to similarity and feature representation are crucial to a CBIR system's effectiveness. Color, shape, texture, and gradient are some essential features that can be utilized to portray an image. Local Binary Pattern (LBP) is a modest and successful texture controller that marks the pixels of an image by controlling the part of every pixel and deciphering the outcome as a binary value. The Local Binary Pattern (LBP) approach is acquainted with grey-level images to characterize color images as the pattern's dimensionality is enhanced. The current study proposes the 'Median Binary Pattern', which incorporates the multichannel decoded Local Binary Pattern (mdLBP) utilized to portray color images. For consolidating LBPs from more than one channel to make the descriptor noise-robust, two structures, specifically adder and decoder-based structures, and a noise-robust binary pattern called the 'Median Binary Pattern'. Compared with existing approaches, the proposed method achieved Average Recovery precision (ARP) and Average Recovery rate (ARR) of 68.1 and 33.55, respectively, with Noise Robust Binary Patterns.