Receiving the most relevant images from image databases is a challenging and critical issue in many applications. Texture is a substantial feature of an image which depicts the spatial behavior of gray-levels in any given neighborhood. Color features uses a variety of color systems and are meaningful to differentiate image segments. Presently, many of the favorable methods for image content description use local descriptors as their starting point with several conducts. The content in an image may appear in some feature descriptor's components more accurately than other components. This paper presents an innovative idea for local image retrieval using a new methodology for feature extraction welding named Blend of Extended Features' Components (BoEFC). The paper shows that an image's content may be described individually by the feature descriptor's components or collectively through the Extended Feature Components (EFC). Retrieval options are attempted using a selection method of Feature Components then the relevant results are collected and ordered according to newly adapted feature similarity measures. The experiments were performed using a general-purpose image database which itself represent a challenge and the INRIA Holiday image database. The experiments was performed by varying the EFCs to compute recall, precision and draw the Precision-Recall (PR) curves which showed increased recall and precision with some components. In addition, calculating mAP and mAR showed increased performance due to the BoEFC blending process.