The rapid growth of multimedia content has created a demand for effective communication systems. Traditional systems treat all pixel values the same while applying compression techniques, eventually significantly losing important information from the image. To address this problem, this paper proposed a saliency-guided semantic communication, where the salient object from each image is separated as Region of interest and Non-Region of Interest, and to make the transmission efficient different compression rates are applied to these regions. The salient and non-salient regions of the image are transmitted and reconstructed at the receiver side. To show the efficacy of the proposed system, numerical simulations are conducted on VOC2012 dataset. The obtained results show the effectiveness of the proposed method in terms of better high Structural Similarity Index Measure, Peak Signal-to-Noise Ratio, lower Mean Square Error and Contrastive Language–Image Pretraining over LTE and 5G channels. Proposed saliency-guided ROI approach delivers a 40% improvement in energy efficiency, and a 37% reduction in average transmitted bits per image over the JPEG baseline transmission.