This paper studies the transmission of progressive images over multiple-input multiple-output (MIMO) channels with inter-cell interference. For the transmission of progressive images, the optimization of jointly allocating various spectral efficiencies and spatial multiplexing rates to a series of numerous packets has been a challenging problem. In addition, if transmit power control is also involved with the optimization for interference channels, the problem becomes more complicated. To address such issues, in this paper, a neural network based optimization method is proposed, where the input layer of the network takes the distortion-rate characteristics of progressive images and the estimates of channel propagation gains between transmitters and receivers. Unsupervised learning is adopted to train the network, where the softmax activation functions of the output layer are exploited to average the expected distortions of the images over the predicted probability distributions of spectral efficiencies and spatial multiplexing rates. The proposed scheme offers significantly improved performance, compared to the baseline method that separately optimizes transmit power and a set of spectral efficiencies and spatial multiplexing rates. It is demonstrated that a neural network is able to analyze and learn the highly nonlinear distortion-rate characteristics of images that are associated with MIMO channel fading effects in the presence of inter-cell interference.