Imperfections in imaging systems and object motion introduce blur in images, leading to an inaccurate analysis. Users are typically unaware of the blurring present in an image, and for accurate analysis, the blur in the images needs to be removed using blind image deblurring. Blind image deblurring serves as a viable method for addressing various sources of blur without prior knowledge of the blur kernel. This paper proposes an edge-aware and K-means cluster-based Manhattan regularization for deblurring (EA-KCMRD). The proposed blind image deblurring algorithm addresses the real-world and synthetic blur by integrating explicit and implicit methods for salient edge selection. The proposed EA-KCMRD introduces a hybrid edge selection framework that combines adaptive Canny edge detection with mutually guided image filtering (explicit strategy) and K-means clustering with Manhattan regularization (implicit strategy). This integration preserves both prominent and structural details, making the kernel estimation process more robust under noise and low-contrast conditions. Furthermore, EA-KCMRD employs a multi-scale framework that supports image deblurring for different image sizes across diverse datasets. Experimental results have validated the feasibility of the proposed method on RealBlur-R, GoPro, HIDE, RS-Blur, and MC-Blur dataset. The proposed EA-KCMRD method has been compared with state-of-the-art techniques using objective metrics and subjective visual perception of deblurred images to demonstrate its effectiveness.