Electrical tomography (ET) is an advanced visualization technique owing to low cost, fast response, non-invasiveness and non-radiation advantages. ET must depend on an imaging algorithm to visually reconstruct all objects, and almost all algorithms with high spatial resolution contain a vital parameter at least. Moreover, the similarity norm plays an important role in the ET imaging process. If they cannot accurately be determined, the ET reconstruction quality is not guaranteed. In this paper, according the clustering mechanism of the ET process, a clustering validity index (CVI) is used to determine both the hyperparameter and similarity norm in the ET algorithm, while CVI originally is used to find the optimal number of clusters among various candidates. The within-cluster and between-cluster distances in CVI are newly defined by not only the grey level but also the neighboring information of any pixel in an ET image. Two representative ET algorithms act for a general framework to validate the proposed method along various hyperparameters and similarity norms. Experimental results show that the proposed method can improve the spatial resolution of the ET image by finding optimal parameter and the similarity norm.