The modern era necessitates efficient smart image retrieval from various image collections. Image retrieval relies heavily on primitive image signatures and their internal features. Image retrieval relies heavily on deep metric learning, which aims to identify semantic similarities between data points in the image for accurate image retrieval procedures. The image shape feature representation was generated using a histogram image processing model. To limit the search space, the image pixel shape-based retrieval procedures are effectively used for image retrieval. The dominant colour, edge and shape descriptor has become a common feature in image processing applications. Because of lighting and other variables, colour in nature can shift slightly. A consistent region of an image is detected and extracted from this consistent zone for an accurate image retrieval strategy by performing image segmentation. The proposed model implements a Related Edge Clustered Pixel Extraction Model with Weighted Feature Vector Set (RECPE-WFVS) for extracting the image content set for searching with the query image for an accurate image retrieval procedure. The proposed model is compared with the traditional Remote Sensing Image Retrieval approach based on Fully Convolutional Network (RSIR-FCN) and Classification Using High-Resolution Remote Sensing Images (CHR-RSI), and the results represent that the proposed model's performance is high.