Speckle noise offers a considerable challenge in the turf of digital image processing, exclusively in color images due to its complicated and unexpected nature. Conventional techniques for removing speckle noise frequently fail to properly reduce noise while maintaining image features. Deep learning methods have demonstrated encouraging performance in a number of image restoration challenges in recent years. In this work, we present an innovative method using a supervised denoising convolutional neural network (SDCNN) for speckle noise reduction in color images. The presented SDCNN successfully distinguishes between noise and actual image information by utilizing deep learning to extract complex patterns and characteristics from noisy input imagery. To improve the network’s competence, the suggested method develops a multi-scale structure that allows the model to make decisions at various dimensions by capturing contextual information at both the local and global levels. In addition, the proposed method utilizes residual learning to ease the training procedure and mitigate the vanishing gradient issue. The suggested SDCNN offers a reliable and effective method for eliminating speckle noise from color images. This makes it potentially useful in fields like underwater imaging, satellite imaging, medical imaging, and other fields where precise image reconstruction is crucial.