Abstarct: Removing blur from color images is a significant challenge in the field of computer vision and image processing. Blurring in images can occur due to various reasons such as camera or subject movement, unfocused camera and atmosphere turbulance. This issue not only reduces the visual quality of images but also affects the performance of image processing and computer vision algorithms. Therefore, reconstructing sharp images from blurred ones is considered an important and practical problem. In this research, our primary objective was to develop an efficient method for deblurring color images using prior knowledge of the images. To this end, a deep neural network with a SqueezeUNet architecture was proposed. This architecture combines the compact features of SqueezeNet with the reconstruction power of UNet, resulting in a significant improvement in reconstructing sharp images from blurred ones. The proposed method is designed to reduce model complexity and optimally use computational resources while maintaining high accuracy in image recovery. To evaluate the performance of the proposed method, the GoPro dataset, which contains 2103 sharp images and 2103 blurred images, was used. The experiments were conducted in a Python 3 environment, and the obtained results demonstrate the model's outstanding performance in image reconstruction. Specifically, the proposed method achieved a PSNR of 33.47 and an SSIM of 0.958, demonstrating the effectiveness of the SqueezeUNet model in reconstructing sharp images from blurred ones.