Fourier Series and Laplacian Noise-Based Quantization Error Compensation for End-to-End Learning-Based Image Compression.