In this letter, based on deep joint source-channel coding (DeepJSCC), we propose a channel adaptive scheme based on entropy model and a subchannel matching method with entropy indication to minimize reconstruction distortion for wireless image transmission over orthogonal frequency division multiplexing (OFDM) channels. Specifically, after an image is compressed and packaged into several OFDM packets, the more critical OFDM packets are mapped to subchannels with higher quality based on estimated channel state information (CSI). In addition, after analyzing the effect of channel signal-to-noise ratio (CSNR) on the parameters of our network model, we achieve the adaptation of a single model to various CSNRs simply by adapting the training strategy, without the need to input CSNR into additionally introduced network. Extensive numerical experiments show that our method achieves state-of-the-art performance among existing DeepJSCC schemes over OFDM channels.