In order to help hepatic surgery planning, an unsupervised method is needed to automate the delineation of liver tumors. Moreover, due to the large amount of images acquired by Computer Tomography Scanner (CT-Scan), the processing has to be fast for a clinical use. Current methods are based on filterings and have the drawback of being time consuming. In this paper, to reach the purpose of speed and quality, we propose a fast unsupervised method which is implementable on an opto-electronical processor. The proposed method is based on the expansion/compression paradigm and combines a multiresolution approach with the principal component analysis (PCA). The multiresolution representation is done by several Gaussian filterings. The compression of the expanded information is then achieved by only keeping the first PCA factorial image. Endly, the object of interest is detected and delineated using the standard valley thresholding technique which is applied to the first factorial image. For the delineation of liver tumors, regions of interest (ROI) containing tumors have been preliminary extracted before applying PCA. Experimental results obtained by the processing of difficult clinical cases show, according to the radiologist experts, that our method is able to efficiently delineate liver tumors. Because Gaussian filterings are time consuming when carried out on a digital processor, we propose to implement them on an optical correlator. Clinical cases have been processed using the resulting opto-electronical processor to show the feasibility of such an implementation.