Light field camera acquires an image set by capturing a scene at several viewpoints. However, it has a trade-off issue between the number of views and the spatial image resolution. To address these problems, this paper presents a light field view synthesis using multiplane image generation. Since multiplane images consisting of color values and transparency of objects are projected onto some planes respectively representing disparity information, it is possible to synthesize and transform images to arbitrary viewpoints. The proposed method constructs the multiplane images using 3D convolutional neural network, and then synthesizes a novel view by blending the multiplane images. For the experiment, the proposed method demonstrates the quality is better than existing methods.